# lightweight reads of the saved outputs
anno  <- read_excel(file.path(res, "annotation/consensus_3628_annotation.xlsx"))
motif <- read_excel(file.path(res, "annotation/consensus_motif_enrichment.xlsx"), sheet = "up_opened")

n_total <- nrow(anno)
n_up    <- sum(anno$direction == "up_in_GFP")
n_down  <- sum(anno$direction == "down_in_GFP")
barry_up <- sum(anno$in_barry & anno$direction == "up_in_GFP", na.rm = TRUE)

# collapse ChIPseeker annotation strings into feature buckets for a summary table
bucket <- function(a) dplyr::case_when(
  grepl("^Promoter", a) ~ "Promoter (≤3 kb)",
  grepl("^Distal",   a) ~ "Distal intergenic",
  grepl("Intron",    a) ~ "Intron",
  grepl("Exon",      a) ~ "Exon",
  grepl("UTR",       a) ~ "5′/3′ UTR",
  grepl("Downstream",a) ~ "Downstream",
  TRUE ~ "Other")
feat_tab <- anno |>
  mutate(feature = bucket(annotation)) |>
  count(feature, name = "peaks") |>
  mutate(percent = round(100 * peaks / sum(peaks), 1)) |>
  arrange(desc(peaks))

1 Summary

We profiled chromatin accessibility (scATAC-seq) in FACS-sorted injured (GFP⁺/ATF3⁺) DRG neurons and compared them against contralateral and wild-type naive controls. Injury predominantly opens chromatin at distal regulatory elements, and the injury-opened regions are strongly enriched for AP-1 (Fos/Jun) transcription-factor motifs: the canonical master-regulator program of nerve injury and axon regeneration.

  • 3627 injury-specific (consensus) peaks: 2852 opened, 775 closed after injury
  • 7760% at distal / intronic elements
  • 588 opened peaks near an injury up-regulated gene (Barry et al. 2023)
  • top enriched TF family: AP-1

2 Injured neuron atlas

643 GFP⁺ neurons over 188,277 peaks (QC uniformly high; all cells retained). They resolve into five known DRG subtypes, dominated by peptidergic nociceptors (PEP).

fig("eda/umap_celltype.png")
UMAP of injured (GFP⁺) neurons by subtype.

UMAP of injured (GFP⁺) neurons by subtype.

Caveat. Heavily PEP-weighted (~68%); NP1/Th are only ~20–30 cells, limiting per-subtype power.

3 Differential accessibility (injury vs control)

GFP⁺ neurons were integrated with both controls (WT n=1,213; Contra n=3,134; GFP n=560) on a shared peak set and tested with a depth-corrected logistic-regression model (GFP = reference; positive = opened after injury). Injury-specific (consensus) peaks are those significant against both controls in the same direction: 3627 peaks (2852 opened, 775 closed).

fig("da/DA_two_contrast_scatter.png")
Effect sizes vs the two controls are tightly concordant (only 5 peaks discordant); the injury signal does not depend on the choice of control.

Effect sizes vs the two controls are tightly concordant (only 5 peaks discordant); the injury signal does not depend on the choice of control.

On the extreme-log2FC cluster. A small detached group shows log2FC ≈ 11; this is a pseudocount ceiling (control accessibility ≈ 0), not a real effect size. These peaks are open in only ~10–17% of GFP cells; we prioritise by accessibility prevalence, not raw log2FC.

4 Genomic annotation of injury peaks

The injury peaks map overwhelmingly to distal and intronic regions rather than promoters : i.e. injury reshapes distal regulatory elements (enhancers).

kable(feat_tab, caption = "Genomic feature distribution of the injury peaks (ChIPseeker, mm10).")
Genomic feature distribution of the injury peaks (ChIPseeker, mm10).
feature peaks percent
Distal intergenic 1601 44.1
Intron 1215 33.5
Promoter (≤3 kb) 462 12.7
Exon 243 6.7
5′/3′ UTR 102 2.8
Downstream 4 0.1 
fig("annotation/consensus_feature_distribution.png")
Genomic feature distribution (ChIPseeker).

Genomic feature distribution (ChIPseeker).

Cross-reference with the transcriptome. Of the 2852 injury-opened peaks, 588 (≈21%) have a nearest gene in the injury up-regulated set from Barry et al. 2023 (bulk RNA-seq of DRG subtypes), linking accessibility to the known injury transcriptional response.

Over-representation test. To quantify this rather than assert it, we tested whether injury up-regulated genes fall near injury-opened peaks more than expected by chance — a hypergeometric test with all ATAC-assignable genes as background and the injury-closed peaks as a negative control.

ora <- readRDS(file.path(res, "annotation/hypergeometric_barry_enrichment.rds"))
ora |>
  transmute(`peak set` = set, `genes near` = near_n, observed = observed_k,
            expected = round(expected, 1), `fold` = fold_enrichment,
            `odds ratio` = odds_ratio, `p (hypergeom.)` = signif(p_hyper, 3)) |>
  kable(caption = "Over-representation of Barry injury up-regulated genes near consensus peaks (background = all ATAC-assignable genes).")
Over-representation of Barry injury up-regulated genes near consensus peaks (background = all ATAC-assignable genes).
peak set genes near observed expected fold odds ratio p (hypergeom.)
injury-OPENED (all-ATAC background) 2219 427 304.4 1.40 1.58 0.00000
injury-CLOSED (negative control) 677 116 92.9 1.25 1.31 0.00612

Injury up-regulated genes are significantly over-represented near injury-opened peaks (427 observed vs 304 expected; odds ratio 1.58, hypergeometric p = 1.15^{-14}). The effect is specific to opened chromatin — the injury-closed negative control is far weaker (odds ratio 1.31, p = 0.00612). The modest fold (~1.4×) is expected: nearest-gene assignment is noisy and distal enhancers need not regulate the literal closest gene, so the true regulatory coupling is diluted.

5 TF motif enrichment → candidate master regulators

Scanning the injury-opened peaks for TF motifs (JASPAR2020, GC-matched background) gives a coherent result: the top motifs are almost entirely the AP-1 (bZIP) family (Fos/Fosl1/2/Fosb and Jun/Junb/Jund dimers, JDP2, BACH2), with Smad2::Smad3 (TGF-β) and NFE2.

motif |>
  transmute(motif = motif.name,
            fold_enrichment = round(fold.enrichment, 2),
            percent_observed = round(percent.observed, 1),
            p.adjust) |>
  head(20) |>
  kable(caption = "Top TF motifs enriched in injury-opened peaks.")
Top TF motifs enriched in injury-opened peaks.
motif fold_enrichment percent_observed p.adjust
FOS::JUNB 4.41 44.3 0
BACH2 4.40 37.1 0
FOSL2::JUN 4.34 48.8 0
FOSL2::JUND 4.31 42.3 0
FOS::JUN 4.31 46.5 0
FOSL2::JUNB 4.28 39.0 0
JDP2 4.27 33.6 0
JUN::JUNB 4.22 39.3 0
JUND 4.21 42.1 0
FOSB::JUNB 4.19 43.9 0
FOSL2 4.16 54.2 0
FOSL1 4.14 48.1 0
FOS::JUND 4.11 53.8 0
FOSL1::JUN 4.09 51.4 0
JUNB 4.05 43.1 0
FOSL1::JUNB 4.04 43.5 0
FOSL1::JUND 3.96 39.2 0
Smad2::Smad3 3.96 57.4 0
NFE2 3.92 37.4 0
FOS 3.90 52.9 0 
fig("annotation/consensus_up_motif_enrichment.png")
Top TF motifs enriched in injury-opened peaks (all adjusted p ≈ 0).

Top TF motifs enriched in injury-opened peaks (all adjusted p ≈ 0).

Interpretation. AP-1 (Fos/Jun) is the canonical master-regulator program of peripheral nerve injury and axon regeneration; its dominance validates the analysis and nominates AP-1 (with TGF-β/Smad input) as the upstream driver of the injured-neuron accessibility program.

6 Next steps

  1. Prioritise injury-opened peaks by accessibility prevalence + distance to injury genes; run the targeted ±500 kb search.
  2. Resolve the AP-1 member(s) driving the program (paired expression + TF footprinting / chromVAR).

–