Mechanistic interpretation of biological tissue growth experiments with a computational model

This paper presents a computational model that simulates the interplay between geometry, mechanics, and stochasticity in tissue growth to enable the quantitative analysis of morphology and the inference of dynamic growth mechanisms from static experimental data, demonstrated through applications in 3D-printed scaffolds and cortical bone formation.

The Gist

The Big Idea: Solving a Biological Mystery with a "Time Machine"

Imagine you walk into a room and see a finished sandcastle. You can see the towers, the moat, and the shells used for decoration. But you don't know how it was built. Did the wind blow the sand there? Did a child build it carefully? Did a wave crash over it halfway through?

This is the problem scientists face with biological tissue growth. When they look at a piece of bone or a patch of healing skin under a microscope, they see the final result (the "sandcastle"). But they can't see the process. They can't easily measure the invisible forces (mechanics) or the specific decisions cells made (biology) that shaped it.

The Solution: The authors built a computational "time machine." Instead of just looking at the finished tissue, they created a computer simulation that acts like a virtual lab. They can build tissue from scratch, cell by cell, and watch how different rules (like "cells get crowded" or "cells get stiff") change the final shape.

---

How the Computer Model Works: The "Living LEGO" Analogy

Think of the tissue interface (the edge where new tissue is growing) as a line of Living LEGO bricks.

1. The Push (Growth): These bricks are constantly trying to push outward, adding new material (like laying down more bricks).
2. The Squeeze (Mechanics): As they push, they bump into their neighbors. If the space is tight, they get squished. The model treats them like springs. If they are squished too hard, they push back. This simulates the physical pressure cells feel in real life.
3. The Randomness (Stochasticity): Life isn't a perfect machine. Sometimes a cell decides to split in two (proliferate), sometimes it decides to stop working and become part of the wall (differentiate), and sometimes it just gives up and leaves (dies). The computer adds a little bit of randomness to these decisions, just like real life.

The magic of this model is that it doesn't just show a blurry blob growing. It tracks every single cell. It remembers exactly where a cell was when it stopped moving and became part of the structure. This allows the scientists to compare their "virtual tissue" side-by-side with real "lab tissue" down to the individual cell.

---

The Two Case Studies: Testing the Model

The authors tested their "time machine" on two very different scenarios to see if it could explain real-world mysteries.

1. The Square Room (Tissue Engineering)

The Scenario: Scientists grew cells inside a tiny, 3D-printed square box.
The Mystery: The cells filled the box, but the final shape wasn't a perfect square; the corners got rounded off. Why?
The Simulation: The team ran the model with different "personality settings" for the cells:

• Setting A: Cells are very stiff (like concrete blocks).
• Setting B: Cells are squishy (like jelly).
• Setting C: Cells mostly split in two vs. cells mostly stop working.

The Discovery:

• They found that if the cells were too stiff, the corners stayed sharp.
• If the cells were squishy, the corners rounded off nicely, matching the real experiments.
• They also discovered that the cells grew best when they mostly split into one new worker and one "retired" cell (asymmetric division), rather than just stopping work immediately.

The Takeaway: The shape of the tissue isn't just about the box it's in; it's about how "squishy" the cells are and how they decide to multiply.

2. The Irregular Cave (Bone Formation)

The Scenario: Inside our bones, there are tiny tunnels (pores) that need to be filled with new bone. These tunnels are often lopsided or asymmetrical.
The Mystery: Why are some tunnels perfectly round, while others are weirdly shaped with thick walls on one side and thin walls on the other?
The Simulation: The team simulated bone growth in a random, bumpy hole. They turned off the "cell splitting" feature (since bone cells usually stop multiplying once they start building) and let the cells just add material and retire.

The Discovery:

• Even with perfectly fair rules (no bias toward one side), the bone grew unevenly!
• Why? Because of random chance. Sometimes a cell on the left side decided to retire a split-second earlier than a cell on the right. That tiny delay meant the right side grew a bit further. Over time, these tiny random differences added up to create a lopsided tunnel.

The Takeaway: You don't need a "bad design" or a "defect" to get an asymmetrical bone. Sometimes, randomness alone is enough to make things look uneven.

---

Why This Matters: The "Fingerprint" of Growth

The most important part of this paper is the idea that tissue leaves a fingerprint.

Just like a detective can look at a broken window and guess how it was hit, scientists can look at the arrangement of cells in a piece of bone and guess what happened during its growth.

• If the cells are lined up a certain way, it tells us the tissue was under pressure.
• If the cells are spaced out randomly, it tells us the growth was driven by chance.

In simple terms: This paper gives scientists a new tool. Instead of guessing how a tissue grew by looking at a static photo, they can now run a simulation, tweak the rules, and see which rules produce a "photo" that matches reality. This helps them understand diseases, design better artificial organs, and figure out how our bodies heal themselves.

Summary

• The Problem: We can see the result of tissue growth, but not the process.
• The Tool: A computer model that builds tissue cell-by-cell, adding physics and randomness.
• The Result: The model proved that cell stiffness shapes tissue corners and that random chance can create asymmetrical bones.
• The Future: This helps us reverse-engineer biology, turning static pictures into dynamic stories of how our bodies build themselves.

Technical Summary

1. Problem Statement

Biological tissue growth is governed by a complex interplay between substrate geometry, mechanobiological processes (mechanical stresses affecting cell fate), and stochastic biological events (proliferation, differentiation, death).

• The Challenge: Disentangling the specific contributions of geometry versus mechanics is experimentally difficult. Mechanical properties are hard to measure in vivo, and tissue samples provide only static "snapshots" in time, obscuring the dynamic history of how the tissue architecture was formed.
• The Gap: While existing mathematical models (phenomenological or continuum mechanics) describe macroscopic growth or tissue deformation, they often fail to generate synthetic tissue compositions that are directly comparable to experimental data at the single-cell resolution. They typically focus on interface evolution rather than the internal distribution and orientation of differentiated cells.

2. Methodology

The authors developed a discrete, stochastic computational model that simulates tissue growth by evolving a one-dimensional chain of cells representing the tissue interface.

A. Mathematical Framework

The model evolves the position of cell boundaries ($\mathbf{r}_i$) based on two velocity components:

1. Mechanical Relaxation ($\mathbf{v}^m_i$): Cells are modeled as mechanical springs. In an overdamped regime (viscous environment), velocity is proportional to the net tangential force resulting from restoring forces between neighbors and reaction forces from the substrate.
2. Tissue Formation ($\mathbf{v}^f_i$): Cells deposit Extracellular Matrix (ECM) at a constant rate ($k_f$), displacing the interface perpendicularly. The velocity magnitude is inversely proportional to local cell length (density).

B. Stochastic Cellular Events

The model incorporates four stochastic events occurring at constant rates:

• Symmetric Proliferation: A cell divides into two active interface cells.
• Asymmetric Proliferation: A cell divides; one remains active, the other differentiates and embeds into the ECM.
• Apoptosis: A cell dies and is removed from the interface.
• Differentiation: An active cell differentiates and embeds into the ECM.

Key Novelty: The framework explicitly records the positions and orientations of differentiated cells (embedded in the ECM) as they are generated. This allows for direct, single-cell resolution comparison between synthetic data and experimental images.

C. Calibration and Case Studies

The model was calibrated using experimental data from two distinct scenarios:

1. Tissue Engineering: MC3T3-E1 osteoblastic cells growing in 3D-printed square porous scaffolds. Parameters were tuned to match differentiated cell density ($\rho$) and ECM formation rates ($k_f$).
2. Bone Formation: Osteoblasts forming bone within cortical pores (Haversian systems). The model simulated the deposition of mineralized tissue without proliferation, focusing on differentiation and apoptosis.

3. Key Contributions

• Synthetic Data Generation: A framework capable of generating synthetic tissue compositions (including cell density, orientation, and interface morphology) that are directly comparable to experimental microscopy data.
• Mechanistic Inference: The ability to infer underlying dynamic rates (e.g., differentiation modes) from static, single-time-point experimental data by matching model outputs to observed tissue composition.
• Disentangling Stochasticity vs. Determinism: A method to distinguish whether observed structural irregularities (e.g., asymmetric pore closure) arise from biological heterogeneity or intrinsic stochastic cellular processes.
• Quantitative Linkage: Establishing a mathematical link between cell stiffness, interface morphology, and cell orientation distributions.

4. Key Results

Case Study 1: Tissue Engineering (3D-Printed Scaffolds)

• Differentiation Mechanisms: By varying the ratio of direct differentiation ($\alpha$) to asymmetric division, the model found that $\alpha = 0.2$ (mostly asymmetric division) produced interface morphologies and cell distributions that best matched experimental observations. Higher $\alpha$ values resulted in rougher, less consistent interfaces.
• Population Dynamics: Simulations with a growing population of tissue-forming cells ($\beta > 0$) better captured the late-stage growth dynamics and pore infill rates observed in experiments compared to constant or shrinking populations.
• Cell Stiffness and Orientation:
  • Stiffness: Intermediate cell stiffness ($k \approx 100$ N/μm) produced tissue interfaces with "rounding" similar to experiments.
  • Orientation: Lower stiffness ($k \approx 5$ N/μm) resulted in cell orientation distributions with reduced anisotropy, which better matched experimental cell orientation data.
  • Implication: The discrepancy between interface shape (requiring higher stiffness) and orientation (requiring lower stiffness) suggests that surface tension (not currently in the model) likely plays a significant role in tissue organization.

Case Study 2: Bone Formation (Cortical Pores)

• Osteocyte Density: The model confirmed that osteocyte density remains consistent with the prescribed constant rate ($\rho$) within newly formed lamellar regions, despite variations in the shape and size of the growth front. This validates the use of static density measurements to infer dynamic differentiation rates.
• Asymmetry: Cortical pores often exhibit asymmetric closure. The model demonstrated that purely stochastic events (random differentiation and apoptosis) are sufficient to generate significant asymmetry in wall thickness ($W.Th_{asym}$), without needing to assume pre-existing biological heterogeneities.
• Density Correlation: Simulations showed higher osteocyte density near thinner walls, matching experimental trends.

5. Significance and Future Directions

• Hypothesis Testing: The model serves as a powerful "in silico" tool to test hypotheses about tissue growth mechanisms that are difficult to isolate experimentally.
• Inverse Problems: It enables the estimation of dynamic biological rates (proliferation, differentiation) from static histological data, effectively turning a "snapshot" into a dynamic reconstruction.
• Future Work: The authors propose extending the model to include:
  • Curvature- and stress-dependent differentiation rules.
  • Surface tension forces acting perpendicular to the interface.
  • Statistical inference techniques (e.g., pair-correlation methods) to rigorously compare synthetic and experimental datasets for parameter estimation.

In conclusion, this work bridges the gap between static experimental observations and dynamic mechanistic understanding, providing a robust computational framework to decode the "fingerprint" of tissue growth processes preserved in biological architecture.