paper
active
2024
paper:biochemical-and-biophysical-research-communications-731-2024-150396

Stress Sharing as cognitive glue for collective intelligences: A computational model of Dukkha as a coordinator for Morphogenesis

ByLakshwin Shreesha·Michael Levin
Cognitive glue & collective intelligenceCollective intelligence & bioengineering morphogenesis

TL;DR

Stress sharing among cells — operationalized as the leakage of a binary homeostatic error signal from stressed cells to their fixed neighbors, inducing temporary tunneling channels for movement — demonstrably accelerates collective morphogenetic problem-solving in a multiscale agent-based model of embryogenesis. In 30×30 grid simulations run over 1000 generations of a genetic algorithm, populations with stress sharing reached maximum phenotypic fitness by generation ~400, while hardwired populations reached only 0.975 fitness and without-sharing populations plateaued at ~0.91 by generation 1000 (p≪0.01 for both comparisons at generation 500). At the 20×20 scale the advantage is even more pronounced: stress-sharing populations solved the morphogenetic target (a binary smiling-face pattern) by generation ~100 while without-sharing populations never reached maximum fitness within 1000 generations. The mechanism operates by extending the cognitive light cone of individual cells: in stress-sharing embryos the radius of influence averages 30 units at developmental step 1 and persists through step 85, versus only 5 units through step 10 in non-sharing embryos. Critically, stress maps generated by the model are not decodable by external observers — similarity between stress gradients and the target pattern either drifts from 0.53 to 0.63 or collapses from 0.48 to 0.25 depending on target geometry, providing no reliable read of the system's goal state. The paper argues this constitutes a primitive form of goal-state privacy and implies that modulators of intercellular stress propagation are tractable biomedical targets for controlling morphogenesis, regeneration, and tissue aging.

What to take away

  1. 1. In 30×30 grid genetic algorithm simulations run for 1000 generations, stress-sharing populations reached maximum phenotypic fitness by generation ~400, compared to a fitness of 0.975 for hardwired and 0.91 for without-sharing populations (p≪0.01).
  2. 2. At a 20×20 grid scale, stress-sharing populations solved the binary smiling-face morphogenetic target by generation ~100, whereas without-sharing populations failed to reach maximum fitness even after 1000 generations of evolution.
  3. 3. The mechanism behind the fitness advantage is long-range cell movement: cells in stress-sharing embryos traversed an average Euclidean distance of ~2500 units per generation (utilizing competency ~4725 swaps), versus ~200 units for without-sharing embryos (competency ~100 swaps).
  4. 4. Stress-sharing embryos exhibit a radius of cellular influence averaging 30 units at developmental step 1, persisting until step 85, compared to a radius of only 5 units terminating at step 10 in non-sharing embryos, quantifying the enlargement of the cognitive light cone.
  5. 5. Stress is implemented as a binary homeostatic error signal; when a stressed cell shares its signal with fixed neighbors within a 3×3 neighborhood, those neighbors temporarily form a tunnel allowing the stressed cell to traverse them without dislodging them from their correct positions.
  6. 6. Despite stress being a direct measure of distance to the morphogenetic setpoint, external observers cannot reliably decode the target pattern from stress maps: similarity to the thumbs-up target drifted from ~0.53 to ~0.63 while similarity to the smiling-face target decreased from ~0.48 to ~0.25 over development.
  7. 7. Hardwired populations (no developmental reorganization, relying on mutation alone) outperformed without-sharing competent populations over long timescales, scoring fitness 0.975 vs. 0.91 at generation 1000 for the 30×30 case, demonstrating that competency without communication is worse than no competency.
  8. 8. The three-condition genetic algorithm framework — (1) stress-sharing, (2) without-sharing but competent, (3) hardwired — can be replicated using a 2D binary-cell lattice initialized by randomly scrambling the target pattern, with selection of the top 10% of phenotypic fitness per generation and repopulation via random cell-pair swaps.
  9. 9. The paper raises the open question of whether HSP90 — already known to occur extracellularly — is a biologically instantiated stress-sharing molecule, and states that reporter assays for stress propagation are currently under bench development to test this prediction.
  10. 10. The model predicts that the progressive reduction of the cognitive light cone during development (from large-scale rearrangements to local fine-tuning) maps onto adult tissue maintenance, suggesting that pharmacological modulators of intercellular stress propagation could be targets for controlling tissue replacement and aging.

Peer brief — for seminar discussion

Shreesha and Levin construct a multiscale agent-based model of morphogenesis embedded within a genetic algorithm to test whether intercellular stress sharing — defined as the leakage of a binary homeostatic error signal from mispositioned cells to their correctly positioned neighbors — confers a quantifiable advantage in reaching anatomical target states. The model represents embryos as 2D binary-cell lattices (tested at 20×20, 30×30, and 50×50) initialized by scrambling a target pattern (a downsampled binary smiling-face image). Cells navigate toward the target via distress-signal-guided swapping; the stress-sharing condition adds a tunnel-creation mechanism in which a moving stressed cell coerces fixed neighbors in a 3×3 neighborhood to temporarily permit passage, preventing dislodgement of correctly placed cells. Three populations are evolved for 1000 generations with top-10%-phenotypic-fitness Darwinian selection: stress-sharing, without-sharing-but-competent, and hardwired (no cell movement). An alternative modeling approach the authors could have used is a continuous-valued reaction-diffusion system rather than a discrete binary lattice, which would better capture graded morphogen gradients but at greater computational cost. The load-bearing finding is that stress-sharing populations reach maximum phenotypic fitness by generation ~400 in 30×30 grids, while hardwired populations achieve only 0.975 and without-sharing populations plateau at ~0.91 by generation 1000 (p≪0.01 for both). At 20×20 the advantage sharpens: stress-sharing solves the problem by generation ~100 while without-sharing never converges. The mechanism is long-range movement — stress-sharing embryos use ~4725 competency units and cells travel ~2500 average Euclidean distance units per generation, versus ~100 competency units and ~200 distance units without sharing. A secondary finding is that stress maps are not externally decodable: similarity between stress gradients and the thumbs-up target moved from ~0.53 to ~0.63 while similarity for the smiling-face target fell from ~0.48 to ~0.25 over development, with no consistent trend. The authors frame this as a primitive form of goal-state privacy analogous to the privacy problem in neural decoding. This implies that a simple leak mechanism — requiring no explicit altruism or inter-agent communication protocol — is sufficient to scale individual-cell homeostatic competency to collective morphogenetic problem-solving, with direct translational relevance to regenerative medicine and synthetic morphogenesis. The paper predicts that HSP90, known to occur extracellularly, is a candidate biological instantiation of the stress-sharing molecule, and that modulators of stress propagation could influence tissue aging by controlling the cognitive light cone radius of adult somatic cells. The most pointed critique a critical reader would raise is the discretization and dimensionality: the 2D binary-cell lattice with only two cell states is a severe abstraction that cannot capture the continuous, mechanically coupled, three-dimensional reality of morphogenesis — cells in vivo do not move through fixed grids, and the lattice topology artificially constrains which movement pathways are even possible, potentially inflating the apparent advantage of the tunnel-creation mechanism. Whether stress sharing retains its advantage in a deformable 3D medium with continuous cell-state values and realistic tissue mechanics remains entirely untested, and the authors acknowledge that evolutionary simulations became computationally intractable beyond 50×50 on a 256-core machine, making biological-scale validation remote.

Findings (4)

Claims (3)

Hypotheses (1)

Questions (1)

Related work— refs + corpus + external arXiv

Cited / in-corpus / arXiv badges show which signals surfaced each row. Multi-source rows weighted higher.

Similar preprints — Semantic Scholar

Cited by (1)

Cross-corpus bridges (12)

same_concept_as · Nomic cosine

External markdown files that talk about the same concept as this entity.

  • aboutblank_kb
    How does stress sharing enable cells to navigate morphogenetic task spaces that would be impossible for uncoordinated cells?questions/how-does-stress-sharing-enable-cells-to-navigate.md0.865
  • aboutblank_kb
    How do cells share stress information to enable larger-scale collective morphogenetic goals?questions/how-do-cells-share-stress-information-to-enable.md0.848
  • aboutblank_kb
    How do individual cells with their own homeostatic competencies coordinate to achieve larger-scale, reliable morphological endpoints?questions/how-do-individual-cells-with-their-own-homeostatic.md0.845
  • aboutblank_kb
    Agent-Based Model Of Morphogenesisframeworks/agent-based-model-of-morphogenesis.md0.842
  • aboutblank_kb
    What mechanisms allow the homeostatic capabilities of individual cells to relate to the more complex homeostatic setpoints of the collective?questions/what-mechanisms-allow-the-homeostatic-capabilities-of-individual.md0.840
  • aboutblank_kb
    How does stress sharing as a conserved signal mechanism enable functional cooperation of cells?questions/how-does-stress-sharing-as-a-conserved-signal.md0.835
  • aboutblank_kb
    What molecular mechanisms implement stress-sharing across cellular populations?questions/what-molecular-mechanisms-implement-stresssharing-across-cellular-populations.md0.820
  • aboutblank_kb
    Lakshwin Shreeshathinkers/lakshwin-shreesha.md0.814
  • aboutblank_kb
    How well does stress sharing work in multiscale systems with emergent behaviors?questions/how-well-does-stress-sharing-work-in-multiscale.md0.814
  • aboutblank_kb
    Stress-Induced Evolutionary Innovationframeworks/stress-induced-evolutionary-innovation.md0.812
  • aboutblank_kb
    Virtual Embryo Modelframeworks/virtual-embryo-model.md0.808
  • aboutblank_kb
    Biochemical and Biophysical Research Communications 731 (2024) 150396papers/cleaned/1-s2.0-S0006291X2400932X-main.md0.803