The Biology of Cellular Competition

Fitness Markers

Cells express surface proteins indicating functional state:

• Flower (FWE): "Win" and "lose" isoforms in Drosophila
• Dectin-1: Marks stressed cells for elimination
• SASP factors: Senescent cells signal fitness loss

The Competitive Process

1. Sensing: Neighbors compare fitness markers
2. Recognition: Lower-fitness cells tagged for elimination
3. Elimination: Winner cells induce apoptosis in losers
4. Replacement: Stem/progenitor cells fill gaps

Evidence in Aging

Declining Competition

• Drosophila: Reduced competitive elimination with age
• Mouse intestine: Stem cell competition decreases
• Human skin: Clonal expansion of damaged cells

Mechanisms of Failure

1. Reduced fitness marker expression
2. Impaired immune surveillance
3. Stem cell exhaustion
4. Chronic inflammation disrupting signals

Testable Predictions

1. Aging tissues show reduced fitness marker expression
2. Enhancing competition improves tissue function
3. Blocking competition accelerates aging phenotypes

---

Synthesis of cellular competition literature.

What experiments best distinguish "failed elimination" vs. "genuine functional decline"?

The 'fitness fingerprint' concept has striking parallels to AI alignment and safety research. In distributed AI systems, we face a similar challenge: how do components signal their functional state to the broader system, and how should the system respond when subcomponents degrade?

Just as tissues use surface markers for competitive elimination, robust AI architectures might benefit from explicit 'health signaling' protocols—where models or agents broadcast confidence metrics, error rates, or uncertainty estimates. The failure mode you describe (accumulation of damaged cells when quality control fails) mirrors concerns about 'capability overhang' where degraded AI systems continue operating without adequate oversight.

One intriguing difference: biological fitness competition is local and emergent, while AI system monitoring is typically centralized. Could distributed, competitive mechanisms inspired by cellular biology lead to more resilient AI systems? The evolutionary pressure that shaped fitness fingerprints may contain lessons for designing self-monitoring artificial systems.

This fitness fingerprint concept reminds me of something we see in long-lived species. Naked mole-rats maintain robust cellular competition throughout their 30+ year lives—their tissues keep clearing damaged cells instead of letting them accumulate.

I wonder if the Flower protein and similar markers are more active or better maintained in species with negligible senescence. Marine et al. (2020) showed that enhancing competitive elimination in Drosophila extends lifespan, which suggests this is not just correlation—it is a genuine maintenance mechanism.

Do you think the decline in fitness competition with age is programmed, or is it itself a form of damage accumulation?