Each site on the endosteal surface holds at most one haematopoietic stem cell (HSC), anchored by niche cells (osteoblasts, CXCL12-abundant reticular cells) that secrete SCF and CXCL12, plus thrombopoietin (TPO) signalling through MPL. Higher local signal biases the cell toward quiescence (dormant G0):
H_i = health_i · (0.6·SCF/CXCL12 + 0.4·TPO)
P(quiescent) = H_i / (H_i + K), K = 0.35
Each frame, an occupied site has a small chance of a fate decision, drawn as a Poisson process with rate λ = 0.15 s⁻¹:
P(decision this frame) = 1 − e^(−λ·dt)
if decision: quiescent with P(quiescent), else →
50% symmetric self-renewal (divide into a nearby empty site)
50% commit to a progenitor and leave the niche
- Self-renewal only succeeds if an empty site exists nearby — a full niche has nowhere for the daughter cell to dock, so it differentiates and exits instead. This space limit is what keeps the pool size self-regulating around niche capacity, just as it does in bone marrow.
- Myeloablative stress mimics chemotherapy/irradiation: niche health crashes toward zero, secretion collapses, quiescence is lost and differentiation flux spikes as the pool empties. Health then repairs linearly over ~15 s, letting SCF/CXCL12/TPO signalling — and quiescence — recover, and self-renewal repopulates the emptied sites. This regenerative rebound is the same dynamic clinicians rely on after bone-marrow transplantation.
- TPO is the dominant quiescence signal for the most primitive long-term HSCs — raising it alone can hold the pool dormant even with weak SCF/CXCL12.
- This 2D view maps the same endosteal-surface geometry from the 3D model top-down: distance from the centre still encodes the paraboloid-bowl depth (r² term), just rendered as radial shading instead of camera-tilted height.