# Proposed Allen Showcase Experiment
## Working title
**Connectotype-Vulnerability Mapping: testing whether mesoscale connectivity motifs predict cell-type-specific transcriptomic stress in the 5xFAD mouse model**
## Core question
Do the wiring motifs captured in the Allen Mouse Brain Connectivity Atlas predict which cell types will show the earliest transcriptomic stress signatures in a neurodegenerative mouse model? Can we pre-identify vulnerable circuits from anatomy alone?
## Why this fits Hongkui Zeng and Brain Science at the Allen Institute
- The experiment directly leverages the **Allen Mouse Brain Connectivity Atlas** — the mesoscale projection matrix that Zeng's group built (Oh et al., *Nature* 2014, PMID: 25470075). Every connection in the matrix is a testable hypothesis about which circuits carry vulnerability.
- It integrates the **Mouse Whole Brain Cell Type Atlas** (Yao et al., *Nature* 2023, PMID: 37656950), which provides the transcriptomic cell-type taxonomy that the experiment will cross-reference against connectivity-derived predictions.
- It uses Allen **Cre driver lines** — the transgenic tools Zeng's lab developed for targeting specific cell populations — to sample predicted-vulnerable and predicted-resilient circuits in parallel.
- The Common Coordinate Framework (CCFv3; Wang et al., *Cell* 2020, PMID: 32386544) anchors all measurements to a shared anatomical reference.
## Proposed biological system
### Mouse model
**5xFAD** (C57BL/6J background) — chosen because:
- Amyloid pathology begins at ~2 months, allowing a 3-month pilot to capture pre-plaque and early-plaque transcriptomic stress.
- Single-cell RNA-seq studies (e.g., Mathys et al., *Nature* 2019, PMID: 31227861; and subsequent 5xFAD-specific scRNA-seq) show that excitatory neuron stress and microglial DAM activation emerge before overt plaque deposition.
- The model is widely available, well-characterized, and compatible with Allen Cre lines on the C57BL/6J background.
### Connectivity-based hypothesis
The Allen Connectivity Atlas provides a normalized projection-strength matrix between ~700 CCF-parcellated brain regions. We hypothesize that:
1. Regions receiving **high convergent input** (many strong projections converging) will show earlier transcriptomic stress than regions receiving divergent or sparse input.
2. Cell types within those high-convergence zones that are **postsynaptic to the strongest-weighted projections** will be the first to show stress signatures.
3. This relationship holds even after controlling for distance from the nearest amyloid plaque.
### Test design
Select **6 circuits** from the connectivity matrix:
- 3 **predicted-vulnerable**: circuits with high convergent-input scores (e.g., retrosplenial cortex receiving hippocampal + thalamic input; subiculum receiving CA1 + entorhinal input; anterior cingulate receiving hippocampal + amygdalar input).
- 3 **predicted-resilient**: circuits with low convergent-input scores matched for region size and cell-type composition (e.g., primary visual cortex, primary motor cortex, somatosensory barrel cortex).
### Cre lines for cell-type targeting
Use Allen Cre lines to isolate specific cell types at each circuit:
- **Rbp4-Cre**: Layer 5 IT/subcerebral projection neurons
- **Tlx3-Cre**: Cortical excitatory neurons (layers 2-5)
- **Pvalb-IRES-Cre**: Parvalbumin+ fast-spiking interneurons
- **Sst-IRES-Cre**: Somatostatin+ interneurons
- **Sim1-Cre**: Thalamic projection neurons (for thalamocortical afferents)
### Experimental approach
1. **Derive connectivity priors**: From the Allen projection matrix, compute convergent-input scores for each CCF region. Rank regions. Select top-3 vulnerable and matched bottom-3 resilient.
2. **Collect tissue**: At 2, 3, and 4 months of age, sacrifice 5xFAD mice and wild-type littermates (n=4 per genotype per timepoint). Microdissect the 6 target regions using CCF-aligned coordinates.
3. **Cell-type-specific snRNA-seq**: Use Cre-dependent nuclear tagging (e.g., Cre-dependent Sun1-sfGFP for INTACT) to isolate nuclei from target cell types, then perform snRNA-seq on sorted populations.
4. **Compute transcriptomic stress**: Quantify early stress signatures — ISR/UPR activation, synaptic gene downregulation, inflammatory milieu — per cell type per region.
5. **Test prediction**: Correlate convergent-input score with transcriptomic stress magnitude across the 6 circuits. The primary endpoint is a significant positive correlation (Spearman rho > 0.5, p < 0.05) between connectivity-predicted vulnerability and observed stress.
## Readouts
- Per-cell-type DEG counts and stress-pathway scores (GSEA for ISR, UPR, oxidative stress)
- Plaque distance (from immunohistochemistry on adjacent sections)
- Cell-type composition shifts per region
- Connectivity-stress correlation across the 6 circuits
## Success criterion
This is a success if the Allen connectivity matrix, used without any pathological information, produces ranked predictions of cell-type vulnerability that significantly correlate with measured transcriptomic stress in 5xFAD mice. This would establish that anatomy contains predictive information about neurodegenerative vulnerability — a result that would directly validate the utility of the Connectivity Atlas for disease research.
## Cost estimate (3-month pilot)
- Mice: 5xFAD × 3 timepoints × 4 per genotype × 2 genotypes = 24 mice, plus Cre-line crosses ≈ 72 mice total (~$3,600)
- snRNA-seq: 6 regions × 5 cell types × 3 timepoints × 2 genotypes = 180 libraries × ~$500/lib = ~$90,000
- IHC and imaging: ~$5,000
- Bioinformatics: ~$10,000
- **Total: ~$110K**
## IIG estimate
If connectivity predicts vulnerability: **IIG = 0.7** (high impact — opens a new avenue for using the Connectivity Atlas in disease research; directly connects two major Allen resources; publishable in a high-impact journal).