Polygenic transcription of OR genes in olfactory progenitors follows a zonal expression pattern.

(A) Schematic illustrating OR zones along the dorsoventral axis, in whole mount views of the MOE (left) and coronal sections (middle). Zone1 (red) is the dorsal-most zone and zone5 (blue) is the ventral-most zone. Zoomed in view of a coronal sections (right) depicts MOE cell populations from different stages of OSN differentiation organized in a pseudostratified fashion from the basal (least differentiated) to apical (most differentiated) layers: HBC, horizontal basal cell; GBC, globose basal cell; INP, immediate neuronal precursor; iOSN, immature olfactory sensory neurons; mOSN, mature olfactory sensory neuron.

(B) t-SNE clustering of single cells from FAC-sorted cell populations with Seurat based on the most variable genes showing the separation of single cells into 6 populations (left panel). The 6 populations were assigned cell identities using expression of known MOE markers[31] (See also Figure S1). Olfactory receptor expression is first detected in INP3 cells (right panel).

(C, D, left panel) t-SNE clustering from FAC-sorted cell populations isolated from dorsal (zones 1-2 in C) or ventral (zones 3-5 in D) MOE microdissections. Cells are colored according to the zonal index of the most highly expressed OR. (C, D, right panel) Plots depicting zonal identities of all the ORs detected in individual INP3 cells from dorsal (C) or ventral (D) microdissections. Y axis shows OR expression in normalized UMI counts for different ORs (separated by black lines). On the X axis each point is a different INP3 cell. ORs are colored according to their zonal index.

(E) OR expression by zonal index in olfactory progenitor INP cells (top) and mOSNs (bottom), determined with bulk RNA-seq, in cells isolated from dorsal-most (zone 1) (left), dorsomedial (zones 2-3) (middle), and ventral-most (zones 4-5) MOE microdissections (right). Note that INP and mOSN cells were FAC-Sorted from the same exact dissection, thus the mOSN OR expression patterns confirm the accuracy of the dissection.

Heterochromatin deposition silences OR genes from lower zones.

(A) Signal tracks of H3K9me3 and H3K79me3 native ChIP-seq from the whole MOE over two representative OR gene clusters that were selected because they harbor ORs from both zone 1 and zone 5. Below the signal track OR genes are colored according to their zonal index: zone1 ORs in red, zone2 ORs in yellow, zone 5 ORs in blue and ORs with unknown zonal index in gray.

(B) H3K9me3 (left) and H3K79me3 (right) native ChIP-seq in the whole MOE. Box plots of read density over OR gene bodies separated by their zonal index depict a pattern of deposition that is high on dorsal-most expressed zone 1 OR genes, progressively decreases with more ventral zonal OR indexes, and is absent on class I ORs.

(C) H3K79me3 native ChIP seq in GBC, INP, iOSN and mOSN populations shows onset of H3K79me3 deposition in INP cells. Each row of the heatmaps shows coverage over an OR gene body (separated into categories by their zonal index). (See also Supplementary Figure S2 for H3K9me3 heatmap).

(D) H3K79me3 native ChIP-seq in mOSNs from zonally dissected MOE. Colored schematics above each heatmap depicts the zone of dissection. (See also Supplementary Figure S2 for H3K9me3 heatmap).

Zonal OR compartmentalization permits OR genes from more ventral zones to be recruited into the OR compartment.

(A) In situ Hi-C contact matrices of a 90Mb region of chromosome 2 that contains 3 large OR gene clusters, depicted with orthogonal boxes under the contact matrices. HiC libraries were prepared from mOSNs FAC-sorted from dorsal-most (zone1), dorsomedial (zone2) and ventral-most (zone 4/5) MOE microdissections. For each contact matrix, we magnify the long-range cis contacts between the large OR cluster that contains ORs from every zone with the OR cluster that contains mostly zone 1-2 ORs (red box on the left) and the OR cluster that contains mostly zone 4-5 ORs (blue box on the right). All cis OR contacts increase from dorsal to ventral OSNs, but zone 4/5 ORs associate with the other ORs only in the most ventral OSNs (as noted by the HiC signal on the blue box).

(B) Heatmaps of average interchromosomal Hi-C contacts between OR genes annotated by their zonal index at 50Kb resolution shows increased trans contacts in OSNs from more ventral zones. OR genes have a similar, intermediate frequency of contacts in the mOSN population where they are expressed, marked with an asterisk.

(C) Dip-C on mOSNs from dorsal and ventral dissected MOE was used to generate haplotype resolved single cell contact matrices and 3D genome structures, as previously described[37].

(D) Analysis of Dip-C contact densities of interchromosomal contacts between ORs genes confirms that ventral mOSNs have increased OR compartment interactions (Wilcoxon rank sum p-value = 9.164e-11).

(E) Single cell heatmaps of pairwise distances between OR genes generated from 3D genome structures in two ventral mOSNs show OR genes from different chromosomes intermingle in a different pattern in the two cells (top). For each cell, heatmaps are sorted by chromosome order and show all OR interactions within 10 particle radii (approximately ~600nm). Representative 3D structures showing the different positioning of three chromosomes (chr19, chr9 and chr2) in the two cells resulting in a different pattern of OR cluster contacts (bottom). See also supplemental Figure S3 for a complete map of all the Dip-C libraries.

NFI paralogue gradients regulate zonal OR expression

(A) Heatmaps showing the expression levels of differentially expressed transcription factors between dorsal, dorsomedial, and ventral cells during the INP and iOSN developmental stages. The shown transcription factors are significantly differentially expressed between dorsal and ventral cells with an adjusted p-value of <0.05, at least a three-fold change in expression, and an expression of at least 15 TPM. A broader list of transcription zonal factors is included in Table S1. The heatmaps is sorted based on expression in ventral cells and a color bar above each heatmap shows the log2 fold change in ventral cells relative to dorsal cells.

(B) Expression levels of NFIA, NFIB, and NFIX at four stages of OSN development in dorsal cells (red), dorsomedial cells (green) and ventral cells (blue).

(C, D) Comparison of OR gene expression in NFI ABX triple knockout and control cells from the whole MOE. NFI transcription factors are deleted either in progenitors (C) using a KRT5-CreER driver or in mOSNs (D) using an Omp-IRES-Cre driver (as described in Figure S4). At the schematic depiction of the experimental strategy at the right of each panel, scissors indicate the differentiation stage of the deletion, and red box the cell type that was FAC-sorted for RNA-seq analysis.

(E) OR expression in NFI ABX triple knockout, Nfi AB double knockout, Nfi X knockout and control mOSNs from ventrally dissected MOE. Knockout was induced in progenitors with KRT5-CreER. Plots show a different pattern of OR gene transcription in the different genotypes. Quantification of differentially expressed ORs for each genotype shown in Figure S4.

Spatial transcriptomics shows dorsalization and homogenization of the MOE upon Nfi A, B, and X deletion.

(A) Schematic depicting our analysis pipeline: Spatial transcriptomics was performed on sections of control (wt) and Nfi ABX cKO MOE. Dimensionality reduction was performed, and spatial spots were clustered based on normalized expression of OR genes.

(B) Heatmaps showing scaled, normalized expression levels of the top 20 highest expressed OR genes per zone in the control (wt) dataset. Unbiased neighborhood analysis and clustering grouped spatial spots into 5 clusters for both control (wt) and cKO MOE (depicted in distinct colors on the top of the heatmaps). Clustering of spatial spots in control (wt) samples reproduces anatomical zones, as spots within each cluster express OR genes with the corresponding zone index (left heatmap). We generated the same clusters in Nfi cKO samples (right heatmap). Although cluster 1 expresses exclusively zone 1 ORs, like in control MOEs, clusters 2-5 exhibit homogenous OR expression, with ventral expansion of zone 2/3 ORs, and reduced representation of zone 4/5 ORs.

(C) Average normalized per-spot expression of the 20 highest expressed OR genes from zone1, zone2, and zone5 is overlaid against H&E histological image of control (wt, top) and Nfi cKO (bottom) MOE sections. Expression of zone1 OR genes is confined to the same anatomical region for both control and Nfi cKO sections. Zone2 OR gene expression is spread to more ventral regions in the Nfi cKO compared to control sections. Expression of zone5 OR genes is detected in fewer spatial spots and at a lower expression level in the knockout sample.

(D) Spatial spots were assigned to a zone based on the highest summed normalized expression of OR genes. Zonal spot assignment of the control (wt) sample visually reproduces known anatomical zones. In the Nfi cKO sample, spots in the dorsal region had the highest expression of Class I and zone1 OR genes, similar to control sample. However, at the rest of the Nfi cKO MOE, most spots have a zone 2 OR identity. Spots assigned ‘None’ did not contain any OR transcripts and were excluded from cluster analysis.

Nfi A, B and X regulate chromatin state and OR compartment formation

(A) Native ChIP-seq for H3K9me3 (top) and H3K79me3 (bottom) in Nfi ABX knockout mOSNs from ventral MOE. Heatmaps of ChIP signal over OR genes show decrease of both histone marks on zone 3-5 index ORs in NfiABX knockout compared to control. Triple Nfi deletion was induced by KRT5-CreER (before OSN differentiation).

(B) HiC in in Nfi ABX knockout mOSNs from ventral MOE. Left: In situ Hi-C contact matrices of a 90Mb region of chromosome 2 from control (top) and NFI ABX triple knockout (bottom) mOSNs, as described in Figure 3A. from ventral MOE show long-range cis interactions between 3 large OR gene clusters. Note that long range cis contacts the zone 4/5 enriched cluster and the mixed cluster dissipate in the triple Nfi cKO (bottom blue box), whereas the contacts of the mixed cluster with the zone 1/2 enriched cluster are preserved (bottom red box). Right: Heatmaps of average interchromosomal Hi-C contacts between OR genes annotated by their zonal index (as described in Figure 3B) in control (top) and triple Nfi cKO (bottom) mOSNs from ventral MOE. Trans contacts between zone 4/5 ORs dissipate, whereas trans contacts between zone 2/3ORs reach intermediate levels detected at “in zone” ORs.

(C) OR expression by zonal index in INP cells isolated from ventral Nfi ABX knockout and control MOE. Nfi ABX INP cells were isolated as described in Figure S4. Log2FC of OR expression in Nfi ABX vs control INP cells, shows a decrease in expression of zone4-5 ORs (right).

Genetic induction of OR transcription in olfactory progenitors determines OR choice in mOSNs

(A) Genetic strategy for transcriptional induction of OR Olfr17 (a zone2 index OR) from its endogenous genomic locus. A genetically modified “tetoOlfr17” allele contains a tetO promoter immediately downstream of the endogenous Olfr17 promoter and an IRES GFP reporter after the coding sequence[44]. In the presence of tTA a high level of tetoOlfr17 is induced from the tetO promoter (top), while in the presence of a high amount of doxycycline tTA is inhibited and transcription is regulated by the endogenous promoter. See also Supplemental Figure S7A for information of the genomic locus of this Olfr17 allele.

(B) tTA driven by the Gng8 promoter is expressed in INP and iOSN cells in the MOE[52]. When Gng8-tTA drives expression of tetoGFP transcription is driven only in progenitor cells located on the basal side of the MOE, where the tTA is expressed (left)[46]. In contrast, when Gng8-tTA drives expression of tetoOlfr17, expression persists in mature OSNs where tTA is no longer present (right). See also Supplemental Figure S7A,B for the sustained and widespread expression of this OR allele after 30 days in DOX treatment and for scRNA-seq data on Gng8 expression during OSN differentiation.

(C) In situ Hi-C in tetoOlfr17 expressing cells shows enriched contacts with interchromosomal Greek Island enhancers over the Olfr17 locus, suggesting tetoOlfr17+ mOSNs are using endogenous mechanisms to sustain Olfr17 expression.

(D) tetoOlfr17 expression in coronal sections of the MOE determined by GFP fluorescence. In the absence of tTA induction tetoOlfr17 expression occurs only in zone2 of the MOE (right); with high tTA induction in progenitor cells expression occurs throughout all zones of the MOE (left); and with low tTA induction in progenitors, due to the addition a low amount of doxycycline, expression occurs in zone2 and spreads dorsally to zone1 (middle). Zoom in shows tetoOlfr17 expression in its native zone2 (i) and ectopic expression in the most ventral zone 5 (ii). Mice on low doxycycline treatment were provided doxycycline at 1ug/ml in water throughout gestation and postnatal life. See Supplemental Figure S7C for the Dox administration protocol.

(E) Quantification of tetoOlfr17 expression relative to a normalized zonal position (illustrated on the left) in tetoOlfr17 (bottom), tetoOlfr17 with Gng8tTA driver (top), and tetoOlfr17 with Gng8-tTA driver on low doxycycline (middle). 6 sections from two replicates were analyzed for tetoOlfr17 with Gng8-tTA; 9 sections from two replicates were analyzed from tetoOlfr17 with Gng8-tTA and low doxycycline; 29 sections from two replicates were analyzed for tetoOlfr17 without tTA. The plot displays a maximum of 1000 cells randomly selected for each condition.

(F) H3K9me3 native ChIP signal over the Olfr17 locus in mOSNs from dorsal MOE (red), dorsomedial MOE (green) and ventral MOE (blue) shows a higher level of heterochromatin in ventral MOE.

(G) Model of OR choice in each zone of the MOE, regulated by the interplay of low level polygenic transcription in INP cells which defines the OR repertoire that can be chosen in each zone, and heterochromatic silencing, which prevents ectopic expression of more dorsal ORs. See also Supplemental Figure S7D, E for ChIP-seq analyses on zone 4/5 INP cells from control and Nfi cKO INP cells.

(related to Figure 1): Experimental strategy for isolating cells at different stages of olfactory sensory neuron development for single cell RNA-seq.

(A) Genetic and experimental strategy for isolating four cell populations at different stages of olfactory sensory neuron development (GBCs, INPs, iOSNs and mOSNs) from the same tissue. A representative FACS plot is shown. GBC, globose basal cell; INP, immediate neuronal precursor; iOSN, immature olfactory sensory neurons; mOSN, mature olfactory sensory neuron.

(B) t-SNE plots of clustering of FAC-sorted cell populations with Seurat based on the most variable genes showing the separation of single cells into 6 populations. Plots show the relationship between the FAC-sorted populations (left) and cell lineage (right).

(C) The 6 populations in (C) were assigned cell identities using expression of known MOE markers, whose expression is shown in the heatmap. Expression for each gene is represented in terms of log2 fold change relative to its average expression.

(related to Figure 2): Progressive accumulation of heterochromatin on OR genes in space and time.

(A) H3K9me3 native ChIP seq in GBC, INP, iOSN and mOSN populations shows a similar onset of deposition to that of H3K79me3. Each row of the heatmaps shows coverage over an OR gene body (separated into categories by their zonal index).

(B) H3K9me3 native ChIP-seq in mOSNs from zonally dissected MOE. Colored schematics above each heatmap depicts the zone of dissection.

(C) H3K79me3 native ChIP-seq in cells at different developmental stages from ventral-most dissected MOE: iOSN (left), mOSN (middle), and a pure population of Olfr1507-expressing OSNs (a zone5 OR) (right). H3K79me3 heterochromatin is absent from zone5 ORs in ventral-most iOSNs and is deposited progressively as the cells mature.

(D) Increased H3K79me3 deposition is correlated with increased interchromosomal interactions between OR gene loci. Heatmap of normalized interchromosomal Hi-C contacts between OR cluster bins. Each bin is ordered by H3K79me3 ChIP signal (gray color bar on top) and annotated according to the zonal index of their resident OR genes: zone1 ORs, red; zone2-3 ORs, green; zone4-5 ORs, blue. OR cluster regions that have higher levels of H3K79me3 and are enriched for zone1 ORs have increased trans Hi-C contacts.

(related to Figure 3): OR compartments are highly variable between single cells but show a consistent difference between cells of different zones.

(A, B) Heatmaps of distances between OR genes computed from Dip-C 3D genome structures show OR genes within 10 particle radii (~600nm) in each dorsal mOSN (A) and ventral mOSN (B). Heatmaps are hierarchically clustered to show intrachromosomal OR gene aggregates.

(related to Figure 4): Nfi ABX deletion in zone5 olfactory epithelium results in a shift in the OR repertoire.

(A) Schematic showing the experimental and genetic strategy for deleting Nfi transcription factors in olfactory progenitor HBC cells using a KRT5-CreER driver (left), and mOSNs using an Omp-IRES-Cre driver (right). Upon induction of Nfi deletion in HBCs with tamoxifen, we chemically ablate the MOE with methimazole, allowing the quiescent HBCs to rebuild the epithelium. After 50 days, OSNs that were produced from the Nfi cKO/control HBCs will be marked red, and isolated by FACS, with the use of a second GFP reporter that marks each OSN differentiation stage.

(B) Olfr1507 (a zone5 OR) immunofluorescence (magenta) in MOE sections of adult Nfi ABX triple cKO mice and littermate control (wt) mice. Nuclei are stained with Dapi (blue). Images were taken in the same ventral location, indicated on the schematic of a coronal section of the MOE.

(C) Olfr17 (a zone2 OR) immunofluorescence (green) in MOE sections of adult NFI ABX triple cKO mice and littermate control mice (wt), shows normal Olfr17 in dorsomedial MOE and ectopic spreading of Olfr17 expression into more ventral zones. Images were taken in the same spots in the MOE, indicated in the schematic. Nuclei are stained with Dapi (blue).

(D) Differential expression analysis of OR genes in the different NFI cKO genotypes. Percentages of significantly upregulated ORs (red) and downregulated ORs (blue) are shown.

(E) Volcano plot showing expression of non-zonal mOSN markers, dorsal mOSN markers and ventral mOSN markers in ventral mOSNs isolated from NFI ABX triple cKO relative to wt MOE. Only 13/200 non-zonal mature markers are significantly downregulated in NFI ABX triple knockout, compared to 90/207 ventral markers significantly downregulated and 29/138 dorsal markers that were significantly upregulated. Blue: significantly differentially expressed genes (p-value < 0.5), red: significantly differentially expressed genes with a > 2-fold change.

(related to Figure 5): Expression of zone2 ORs genes spreads ventrally in Nfi ABX knockout MOE.

(A) Normalized expression per spatial spot of the zone2 index ORs Olfr309 and Olfr17 overlaid against histological H&E tissue image of wt and Nfi ABX knockout MOE. Olfr309 is the highest expressed zone2 OR gene in the wt dataset. Compared to wt, expression of Olfr309 and Olfr17 in the Nfi ABX knockout sample is spread to more ventral areas of the MOE.

(B) Normalized expression per spatial spot of the Olfr1507, the most highly expressed zone5 OR gene in the wt MOE dataset. As expected, Olfr1507 is expressed in the ventral zone 5 anatomical region in the wt dataset. Expression of Olfr1507 in the Nfi ABX knockout is lower than in wt but is localized to the same region of the MOE.

(related to Figure 6): Ventral Nfi ABX cKO cells closely resemble dorsomedial cells in chromatin state and compartment formation.

(A) Heatmap with side-by-side comparison of H3K79me3 ChIP-seq signal over OR genes in control (wt) dorsal, dorsomedial, and ventral mOSNs compared to ventral Nfi ABX cKO mOSNs. Ventral Nfi ABX cKO mOSNs have a chromatin state most similar to that of control (wt) dorsomedial mOSNs.

(B) Side by side heatmaps of average interchromosomal Hi-C contacts between OR genes of different zonal indexes show the nuclear OR interactome in ventral Nfi ABX cKO OSNs is most similar to that of control (wt) dorso-medial mOSNs, especially in regards of the intermediate HiC contacts of the zone 2-3 ORs.

(related to Figure 7): Hijacking OR gene choice by inducing OR transcription at the stage of polygenic OR transcription.

(A) Schematic showing that the tetO promoter and the iresGFP tag have been inserted in the endogenous Olfr17 gene locus in Chromosome 7 (top). Olfr17 (red line) resides in a large OR gene cluster and is surrounded by other OR genes, depicted by black lines. Early induction of tetO-Olfr17iresGFP by Gng8-tTA results in tTA-independent expression as GFP signal is expressed most OSNs 35 days after adult mice were placed in high doxycycline diet (bottom).

(B-C) Schematic of the high doxycycline feeding protocol used to confirm tTA-independent tetO-Olfr17iresGFP expression in mOSNs (B). Adult mice (8week old mice) were placed in high-dox diet for 35 days, and their OSNs continue to express GFP (C), excluding the possibility that tTA traces could support Olfr17 transcription. Note the distinction from the low-dox diet described in Figure 7, where mice were exposed in utero to doxycycline and never had a chance to induce Olfr17 expression at high levels.

(D) t-SNE clustering of single cells into 6 populations (as described in Figure S1) shows Gng8 expression in INP and iOSN cell types but not in mOSNs, consistent with previous reports.

(E, F) H3K9me3 native ChIP in ventral INP cells isolated from Nfi ABX cKO and control (wt) littermate mice (as described in Figure S4). Heatmap of ChIP signal (E) shows decreased heterochromatin on “in zone” OR genes at the onset OR choice in control (zone 4-5 ORs) and triple Nfi cKO (zone 2-3) INPs. Overlay of ChIP signal tracks of over an OR cluster (D, top) shows variable changes between control (wt) and Nfi ABX cKO heterochromatin levels in INPs throughout the cluster. OR genes are colored according to their zonal index. Log2 fold change in expression between ventral Nfi ABX knockout and control (wt) mOSNs for each OR in the cluster (D, bottom). Note that zone 2 OR genes ectopically chosen in ventral cKO OSNs have decreased heterochromatin in ventral cKO INPs.