Radial niche shell profile

Cell-type composition in concentric distance shells around plaques

Published

March 24, 2026

Rationale

Previous notebooks characterise the niche composition within a fixed 50 µm radius around plaques. Here we extend this to concentric shells to ask whether the possible interaction microglia-astrocytes-oligos has a spatial architecture.

Shells (distance to nearest plaque):

Shell	Range (µm)
1	0 – 25
2	25 – 50
3	50 – 100
4	100 – 200
5	> 200

Analysis is run separately for CAA and parenchymal plaques, and separately for Adu and IgG treatment groups.

Setup and data loading

Code

source("R/setup.R")
library(patchwork)

data <- load_slides_and_distances("full")
slide1 <- data$slide1; slide2 <- data$slide2
cell_distances_all <- data$distances_all

# Add cell-type and sample metadata to the distance table
meta_lookup <- bind_rows(
  slide1@meta.data |> as_tibble(rownames = "cell") |>
    select(cell, annotatedclusters),
  slide2@meta.data |> as_tibble(rownames = "cell") |>
    select(cell, annotatedclusters)
)

cells_with_meta <- cell_distances_all |>
  left_join(meta_lookup, by = "cell")

#############################
## Cell-type abundance check
#############################
cells_with_meta |>
  count(annotatedclusters, Treatment.Group, name = "n") |>
  arrange(desc(n)) |>
  kbl(caption = "Total cells per annotated cluster and treatment") |>
  kable_styling("striped", full_width = FALSE)

Total cells per annotated cluster and treatment
annotatedclusters	Treatment.Group	n
Oligodendrocyte	Adu	36516
Oligodendrocyte	IgG	33808
Astrocyte	Adu	25322
Astrocyte	IgG	23167
Glutamatergic Neuron 1	Adu	20693
Microglia	Adu	19614
Glutamatergic Neuron 1	IgG	18728
Glutamatergic Neuron 2	IgG	17705
Glutamatergic Neuron 2	Adu	16630
Microglia	IgG	15185
Endothelial	Adu	13324
Endothelial	IgG	12259
GABAergic Neuron 1	IgG	10955
GABAergic Neuron 1	Adu	10220
GABAergic Neuron 2	Adu	7529
GABAergic Neuron 2	IgG	7497
NA	IgG	6333
OPC	Adu	5352
NA	Adu	5138
OPC	IgG	4526
Glutamatergic Neuron 3	Adu	4239
Fibroblast	IgG	3263
Glutamatergic Neuron 3	IgG	3244
Pericyte	Adu	3213
Fibroblast	Adu	2916
Pericyte	IgG	2714
GABAergic Neuron 3	Adu	2585
GABAergic Neuron 3	IgG	2489
CP	IgG	2444
Glutamatergic Neuron 4	Adu	2085
GABAergic Neuron 4	IgG	2083
Glutamatergic Neuron 4	IgG	2070
GABAergic Neuron 4	Adu	2008
Ependymal	Adu	1929
CP	Adu	1685
Ependymal	IgG	1675
VSMC	IgG	1296
VSMC	Adu	1240
Glutamatergic Neuron 5	Adu	1092
Glutamatergic Neuron 5	IgG	902
BAM	Adu	819
BAM	IgG	658
T Cell	Adu	392
VLMC	IgG	375
VLMC	Adu	283
T Cell	IgG	199

Code

###################
# Shell assignment
###################

shell_breaks <- c(0, 25, 50, 100, 200, Inf)
shell_labels <- c("0-25", "25-50", "50-100", "100-200", ">200")

cells_shelled <- cells_with_meta |>
  mutate(
    shell_caa   = cut(dist_caa,   breaks = shell_breaks, labels = shell_labels,
                      right = FALSE, include.lowest = TRUE),
    shell_paren = cut(dist_paren, breaks = shell_breaks, labels = shell_labels,
                      right = FALSE, include.lowest = TRUE)
  )

Compute shell proportions

For each brain (sample_ID), plaque type, and distance shell, we compute the proportion of each cell type among all cells in that shell. Proportions are then averaged across the 6 brains for the group-level profile, and separately within Adu and IgG groups.

Code

compute_shell_props <- function(shelled_df, plaque_type_col, shell_col) {
  shelled_df |>
    rename(shell = !!sym(shell_col)) |>
    count(sample_ID, Treatment.Group, shell, annotatedclusters, name = "n") |>
    group_by(sample_ID, shell) |>
    mutate(prop = n / sum(n)) |>
    ungroup()
}

props_caa   <- compute_shell_props(cells_shelled, "dist_caa",   "shell_caa")
props_paren <- compute_shell_props(cells_shelled, "dist_paren",  "shell_paren")

###########################################################
# Mean proportion per shell × cell type × treatment group
###########################################################

summarise_props <- function(props_df, plaque_label) {
  props_df |>
    group_by(Treatment.Group, shell, annotatedclusters) |>
    summarise(
      mean_prop = mean(prop, na.rm = TRUE),
      se_prop   = sd(prop, na.rm = TRUE) / sqrt(n()),
      .groups   = "drop"
    ) |>
    mutate(plaque_type = plaque_label)
}

summary_caa   <- summarise_props(props_caa,   "CAA")
summary_paren <- summarise_props(props_paren, "Parenchymal")
summary_all   <- bind_rows(summary_caa, summary_paren)

Radial profile plots

All cell types — area plot by shell

Code

ggplot(summary_all, aes(x = shell, y = mean_prop, fill = annotatedclusters)) +
  geom_col(position = "fill", colour = "white", linewidth = 0.2) +
  facet_grid(Treatment.Group ~ plaque_type) +
  scale_y_continuous(labels = scales::percent) +
  labs(
    title = "Radial niche composition — all cell types",
    x     = "Distance shell (µm from plaque)",
    y     = "Mean cell-type proportion",
    fill  = "Cell type"
  ) +
  theme_minimal(base_size = 13) +
  theme(
    axis.text.x = element_text(angle = 30, hjust = 1),
    legend.position = "right"
  )

Cell-type composition in each distance shell. Each bar shows the mean proportion of each cell type among all cells in that shell, averaged across all 6 brains. Left: CAA plaques; Right: Parenchymal plaques. Top: Adu; Bottom: IgG.

Changes in cell-type composition across distance shells

no treatment effect

Code

## Identify cell-types of interest

ct_astro <- "Astrocyte"
ct_oligo <- "Oligodendrocyte"
ct_micro <- "Microglia"

message("Microglia:        ", paste(ct_micro, collapse = ", "))
message("Astrocytes:       ", paste(ct_astro, collapse = ", "))
message("Oligodendrocytes: ", paste(ct_oligo, collapse = ", "))

glial_types <- c(ct_micro, ct_astro, ct_oligo)
glial_labels <- setNames(
  rep(c("Microglia", "Astrocytes", "Oligodendrocytes"),
      c(length(ct_micro), length(ct_astro), length(ct_oligo))),
  glial_types
)

Code

# Recompute proportions with collapsed glial categories
cells_glial <- cells_shelled |>
  filter(annotatedclusters %in% glial_types) |>
  mutate(glia_type = glial_labels[as.character(annotatedclusters)])

props_glial_caa <- cells_glial |>
  rename(shell = shell_caa) |>
  count(sample_ID, Treatment.Group, shell, glia_type, name = "n") |>
  # Denominator: ALL cells in this shell (not just glia), to show relative abundance
  left_join(
    cells_shelled |>
      rename(shell = shell_caa) |>
      count(sample_ID, shell, name = "n_total"),
    by = c("sample_ID", "shell")
  ) |>
  mutate(prop = n / n_total) |>
  group_by(Treatment.Group, shell, glia_type) |>
  summarise(mean_prop = mean(prop, na.rm = TRUE),
            se_prop   = sd(prop, na.rm = TRUE) / sqrt(n()),
            .groups   = "drop") |>
  mutate(plaque_type = "CAA")

props_glial_paren <- cells_glial |>
  rename(shell = shell_paren) |>
  count(sample_ID, Treatment.Group, shell, glia_type, name = "n") |>
  left_join(
    cells_shelled |>
      rename(shell = shell_paren) |>
      count(sample_ID, shell, name = "n_total"),
    by = c("sample_ID", "shell")
  ) |>
  mutate(prop = n / n_total) |>
  group_by(Treatment.Group, shell, glia_type) |>
  summarise(mean_prop = mean(prop, na.rm = TRUE),
            se_prop   = sd(prop, na.rm = TRUE) / sqrt(n()),
            .groups   = "drop") |>
  mutate(plaque_type = "Parenchymal")

props_glial <- bind_rows(props_glial_caa, props_glial_paren) |>
  mutate(glia_type = factor(glia_type, levels = c("Microglia", "Astrocytes", "Oligodendrocytes")))

Code

glial_colours <- c(
  "Microglia"        = "#E41A1C",
  "Astrocytes"       = "#FF7F00",
  "Oligodendrocytes" = "#377EB8"
)

ggplot(props_glial,
       aes(x = shell, y = mean_prop,
           colour = glia_type, group = glia_type)) +
  geom_line(linewidth = 1.2) +
  geom_point(size = 3) +
  geom_errorbar(aes(ymin = mean_prop - se_prop, ymax = mean_prop + se_prop),
                width = 0.2) +
  facet_grid(Treatment.Group ~ plaque_type) +
  scale_colour_manual(values = glial_colours) +
  scale_y_continuous(labels = scales::percent) +
  labs(
    title  = "Glial radial profile — three-layer relay test",
    x      = "Distance shell (µm from plaque)",
    y      = "Mean proportion of all cells in shell",
    colour = "Glial cell type"
  ) +
  theme_minimal(base_size = 14) +
  theme(
    axis.text.x   = element_text(angle = 30, hjust = 1),
    legend.position = "bottom",
    strip.text    = element_text(size = 13)
  )

Proportion of each glial cell type (relative to all cells in that shell) across distance shells. Points show the mean across 6 brains; error bars are ±1 SE. If the three-layer relay holds, microglia should peak in the innermost shell (0–25 µm), with astrocytes and oligodendrocytes showing broader or more distal peaks.

Adu vs IgG comparison

Code

ggplot(props_glial,
       aes(x = shell, y = mean_prop,
           colour = glia_type, linetype = Treatment.Group, group = interaction(glia_type, Treatment.Group))) +
  geom_line(linewidth = 1.1) +
  geom_point(size = 2.5) +
  facet_wrap(~ plaque_type, ncol = 2) +
  scale_colour_manual(values = glial_colours) +
  scale_linetype_manual(values = c("Adu" = "solid", "IgG" = "dashed")) +
  scale_y_continuous(labels = scales::percent) +
  labs(
    title    = "Glial radial profile by treatment",
    x        = "Distance shell (µm from plaque)",
    y        = "Mean proportion of all cells in shell",
    colour   = "Glial cell type",
    linetype = "Treatment"
  ) +
  theme_minimal(base_size = 14) +
  theme(
    axis.text.x   = element_text(angle = 30, hjust = 1),
    legend.position = "bottom"
  )

Glial radial profiles split by treatment. Solid = Adu, dashed = IgG. Divergence between treatment lines within a shell suggests treatment-dependent spatial reorganization.

Summary table: peak shells per cell type

Code

# For each glia type, plaque type, treatment: which shell has the maximum mean_prop?
props_glial |>
  group_by(glia_type, plaque_type, Treatment.Group) |>
  slice_max(mean_prop, n = 1) |>
  ungroup() |>
  select(glia_type, plaque_type, Treatment.Group, peak_shell = shell, peak_prop = mean_prop) |>
  mutate(peak_prop = scales::percent(peak_prop, accuracy = 0.1)) |>
  arrange(plaque_type, glia_type, Treatment.Group) |>
  kbl(caption = "Peak shell for each glial cell type by plaque type and treatment") |>
  kable_styling("striped", full_width = FALSE)

Peak shell for each glial cell type by plaque type and treatment
glia_type	plaque_type	Treatment.Group	peak_shell	peak_prop
Microglia	CAA	Adu	>200	11.1%
Microglia	CAA	IgG	>200	9.1%
Astrocytes	CAA	Adu	25-50	19.2%
Astrocytes	CAA	IgG	25-50	17.7%
Oligodendrocytes	CAA	Adu	>200	21.6%
Oligodendrocytes	CAA	IgG	>200	21.4%
Microglia	Parenchymal	Adu	0-25	27.3%
Microglia	Parenchymal	IgG	0-25	20.3%
Astrocytes	Parenchymal	Adu	>200	16.9%
Astrocytes	Parenchymal	IgG	>200	15.6%
Oligodendrocytes	Parenchymal	Adu	25-50	23.0%
Oligodendrocytes	Parenchymal	IgG	25-50	21.8%