Paracrine Control of Nrf2: Cell-Specific Knockouts and Single-Cell Sequencing to Test Whether Oxidative-Stress Defense Is a Conversation Between Cells

Nrf2 is a transcription factor that switches on a cell's antioxidant defense program. In healthy conditions a protein called Keap1 constantly tags Nrf2 for destruction; when reactive oxygen species or electrophiles accumulate, that tagging fails, Nrf2 accumulates, and the cell turns on genes like HMOX1, NQO1, and GCLC. Essentially everything we know about this pathway comes from treating it as each cell's private business: a cell senses its own chemical stress and mounts its own defense. But cells in a tissue are not isolated. Neighboring cells continuously exchange soluble, secreted molecules — paracrine signaling — including cytokines, growth factors, extracellular vesicles, and small metabolites. Whether one cell's Nrf2 state changes what it tells its neighbors, and whether those messages in turn set Nrf2 in the neighbors, is largely unmeasured, because the standard tools blur it: pharmacological activators hit every cell at once, bulk assays average over populations, and ordinary co-culture cannot separate diffusible signals from direct cell contact.

The proposal here is to make that non-cell-autonomous layer measurable by combining three tools that are individually mature but have not been put together for this question: conditional gene knockout (Cre-lox, where a enzyme called Cre removes a specific gene only in chosen cell types), paracrine decoupling assays (Transwell porous-membrane co-culture and conditioned-medium transfer, which let soluble signals pass while blocking physical contact), and single-cell RNA sequencing (scRNA-seq, which reads the transcriptome of each cell individually so donor and recipient responses can be separated). The study of secreted mediators acting across separated cells goes back at least to 1960 矢野 武男 et al., but that lineage of work measured secreted factors and gross phenotype, never per-cell pathway state in genetically defined sender and receiver populations. One caveat I state up front: the research file supplied to me was truncated before its sections on intersection prior work and benchmark results, so I could not confirm on Lacuna what the closest existing work at this intersection is, and I found no reported quantitative results to anchor baselines. Every novelty claim below is therefore provisional and will be re-checked against a complete literature search before committing resources.

Core Claim

Nrf2 pathway activation is not purely cell-autonomous: deleting or constitutively activating Nrf2 in one cell population reshapes the set of molecules it secretes, and those secreted changes measurably shift Nrf2 target-gene expression in genetically unperturbed neighbor cells. This cross-cell control can be detected, quantified, and causally validated with cell-type-specific knockout, contact-blocked co-culture, and per-cell transcriptomics.

Research Program

The program has three modules. First, a genetic perturbation module builds two donor lines with Cre-lox conditional alleles: donors lacking Nrf2 (Nfe2l2 floxed, so they cannot mount the antioxidant program) and donors lacking Keap1 (so Nrf2 is constitutively active). Recipients in all experiments are genetically wild type. Second, a paracrine decoupling module cultures donors and recipients in Transwell inserts with 0.4 µm membranes, which block cell migration and contact while letting proteins, vesicles, and small molecules diffuse; parallel conditioned-medium transfers — donor supernatant filtered and applied to recipient cultures — establish that any observed effect is carried by diffusible chemistry rather than residual contact. Third, a measurement and inference module runs scRNA-seq on both compartments across a time course, with sample hashing to label which compartment each cell came from. Per-cell Nrf2 activity is scored as the average expression of an a-priori Nrf2 target-gene module (HMOX1, NQO1, GCLC, GCLM, TXNRD1 and related genes), and ligand-receptor inference pairs donor-side differential secreted-factor expression with recipient-side receptor and downstream target changes. A key output is the relay hypothesis test: whether recipient cells that receive a first-wave paracrine signal then turn on their own distinct secreted program, which neither monoculture secretome would contain.

Proposed Experiments

Experiment 1 asks whether donor Nrf2 state changes recipients at all. Keap1-knockout donors, Nrf2-knockout donors, and control donors are co-cultured with wild-type recipients across a time course; the primary readout is the distribution of per-cell Nrf2 module scores in recipients, compared between donor genotypes. The metric of record is the difference in the fraction of recipient cells above a pre-registered Nrf2-score threshold, plus the effect size (standardized mean difference of module scores) between conditions. No reported result found on Lacuna for expected effect sizes of cross-cell Nrf2 modulation, so the time course is designed with dense early sampling to catch transient effects.

Experiment 2 is the reciprocal test: whether Nrf2 in the recipient is required to interpret paracrine stress cues. Wild-type or stressed donors (for example, donors carrying an oxidative burden) send conditioned medium to recipients that are themselves Nrf2-competent or Nrf2-deficient; if recipient Nrf2 is required, Nrf2-null recipients should show altered downstream transcriptional responses to the same medium.

Experiment 3 is causal validation of the inferred mediators. The top-ranked candidate ligands from the ligand-receiver inference are perturbed with neutralizing antibodies or decoy receptors in the Transwell system, and recombinant versions are applied alone to naive recipients. The prediction metric is precision at k: of the top k predicted mediators, the fraction whose perturbation significantly moves recipient Nrf2 module scores. Targeted protein measurements (ELISA or multiplex bead arrays) of the validated candidates in donor supernatants close the loop between RNA inference and actual secreted protein.

Baselines

The naive reference point is the cell-autonomous response: each genotype or condition cultured alone and scored for Nrf2 activation after direct electrophile stimulation. This is what the field implicitly assumes captures the pathway, so any cross-cell effect must be shown on top of it, using the same cells, media, time points, and sequencing budget. No reported result found on Lacuna for the magnitude of these monoculture responses. The strongest existing-method competitor is the standard bulk readout: whole-co-culture bulk RNA-seq or an antioxidant-response-element luciferase reporter run on the identical Transwell cultures. It competes because it is cheaper, better established, and directly sensitive to Nrf2 activity; it will be run on the same wells, same time points, and an equivalent total sequencing-depth budget. The proposal beats it only if it resolves effects the bulk assay mixes away — specifically, opposite-signed donor and recipient shifts that cancel in bulk, or cell-type-specific responses within the recipient compartment — and those split effects then validate under antibody perturbation in Experiment 3. If bulk readouts capture everything scRNA-seq finds, the added cost is not justified and I would report that. The key-ingredient ablation is scRNA-seq of co-cultures with a global pharmacological Nrf2 activator applied to everyone instead of the conditional knockout. If that arm reproduces the knockout arm's recipient responses, then cell-type-specific genetics is not the necessary ingredient and the simpler perturbation suffices for future screens; the knockout design is vindicated only if the activator arm gives a qualitatively different (or absent) recipient pattern.

Why not just…?

Why not just apply a pharmacological Nrf2 activator to the co-culture? Because an activator like sulforaphane reaches every cell simultaneously, so donor and recipient roles are confounded: you cannot tell whether a recipient's Nrf2 shift came from its own drug exposure or from a neighbor's changed secretome. That said, this is a genuine threat to the design's necessity, which is why the activator arm is built in as the key ablation rather than argued away. If it matches the genetic results, the honest conclusion is that the cheaper tool is adequate for mapping, and the knockouts remain only for causal assignment.

Why not just measure the secretome by mass spectrometry or ELISA? Secretome profiling would find the mediators, and the proposal includes targeted protein measurements for exactly that reason. But protein catalogs alone cannot say which recipient cell types respond, cannot detect the synergistic effects that appear only in co-culture and in neither monoculture's secretome, and cannot see relay signaling — the secondary secreted wave from recipient cells. The single-cell transcriptome is what assigns direction and cell identity; the protein assays are validation, not the primary instrument.

Why not just mine existing single-cell atlases of stressed tissues? Public datasets show Nrf2 target-gene heterogeneity across cell types, but they are observational: genotypes are unbalanced, contact-mediated and diffusible signals are tangled in vivo, and there is no perturbation to separate cause from correlation. The Transwell plus conditional-knockout design exists precisely to impose the control that mining cannot.

Decisive Checks

The program stands or falls on three outcomes: recipient Nrf2 module scores shift with donor genotype in contact-blocked co-culture and reproduce under pure conditioned-medium transfer; neutralizing a predicted mediator abolishes the shift while the recombinant ligand phenocopies it; and at least one relay or synergy effect is found that no monoculture secretome contains. Any one of these failing sharpens the field's picture too — it would bound how far Nrf2's cell-autonomous framing actually reaches — but all three succeeding would establish that oxidative-stress defense is, in part, a conversation between cells.