A mesenchymal stem cell placed on a flat plastic dish will mostly stay a stem cell. Place the same cell on a surface covered in 120-nanometre pits arranged in a slightly disordered grid, and many cells begin depositing bone mineral. No growth factors were added and no drugs were used. The only variable was shape, at a scale roughly a thousand times smaller than the cell itself.

This is one of the more surprising findings in biomaterials research: cells read surfaces. They probe the topography beneath them with molecular machinery fine enough to register features only tens of nanometres across. They then translate that geometry into decisions about where to attach, which way to move, and what to become.

For materials engineers, this reframes the substrate. A surface is no longer just a passive scaffold. It is an instruction set written in geometry. Understanding how cells decode that instruction set, from single molecular bonds up to whole-tissue organisation, opens a route to implants and scaffolds that guide biology physically rather than chemically.

Cells attach to surfaces through integrins, transmembrane receptors that bind short peptide motifs such as RGD in adsorbed proteins like fibronectin. A single integrin bond is weak and short-lived. Useful adhesion requires clustering: dozens to hundreds of integrins gathering into complexes that connect to the actin cytoskeleton through proteins such as talin, vinculin, and paxillin.

Nanoscale geometry governs whether that clustering can happen. Researchers placed gold nanodots at precise spacings on an otherwise non-adhesive background, with each dot carrying a single ligand. When the spacing was below roughly 58 nanometres, cells spread and formed stable adhesions. Above about 70 nanometres, adhesions failed to mature and cells struggled to spread. The ligand chemistry was identical across samples. Only the spacing changed, and a difference of a few tens of nanometres switched the behaviour.

Maturation is a mechanical process. Nascent adhesions smaller than a micrometre are pulled on by actomyosin contraction. Under that tension, talin partially unfolds and exposes hidden binding sites for vinculin, which recruits more actin and reinforces the link. Adhesions that can bear load grow into elongated focal adhesions several micrometres long, while those that cannot bear load disassemble. Topography that restricts where integrins can cluster, through pits, pillars, or gaps in ligand presentation, limits how large adhesions can grow. That in turn limits how much tension the cell can generate.

That tension propagates inward. Cytoskeletal stress activates RhoA/ROCK signalling, deforms the nucleus through LINC complexes, and moves the transcriptional regulators YAP and TAZ between the cytoplasm and the nucleus. The adhesion site acts as a transducer: geometry goes in, mechanical load passes through, and gene expression comes out. This pathway converts a static physical feature into a persistent biochemical signal.

Takeaway

A focal adhesion works like a strain gauge built from proteins. Control the nanoscale space where it can grow, and you control how much force the cell believes its environment can bear.

In 1907, Ross Harrison observed embryonic cells crawling along spider-web threads and following their path. Paul Weiss later named the phenomenon contact guidance: cells align with and migrate along anisotropic, or direction-dependent, topographical features. Nanofabrication has since revealed just how small those features can be.

On substrates patterned with parallel grooves and ridges, many cell types elongate along the groove axis, including fibroblasts, epithelial cells, neurons, and smooth muscle cells. Corneal epithelial cells have been shown to align to ridges around 70 nanometres wide, and some studies report responses to grooves only tens of nanometres deep. Groove depth often matters more than width. Deeper channels produce stronger alignment, up to the point where cells bridge across ridges instead of conforming to them.

The mechanism ties back to adhesions. Filopodia, thin protrusions at the cell's leading edge, probe the surface and stabilise preferentially along ridges where continuous adhesion is possible. Focal adhesions form and elongate parallel to the pattern, and actin stress fibres align with them. Because adhesions running across the grooves are constrained, the cell's traction forces become anisotropic. Polarisation and directed migration follow from that imbalance in force.

This matters because native tissue is itself nanopatterned. Collagen fibrils in tendon, ligament, and cornea are tens to hundreds of nanometres in diameter, are highly aligned, and carry a banding pattern that repeats every 67 nanometres. Engineered grooved scaffolds and electrospun aligned nanofibres exploit the same cue. They are used to guide regenerating nerves across injury gaps and to organise cells into tissue with directional mechanical properties.

Takeaway

Contact guidance turns a surface into a track. Anisotropy in the substrate becomes anisotropy in force, and anisotropy in force becomes direction.

The most consequential demonstration came in 2007. Researchers patterned polymer surfaces with pits 120 nanometres in diameter and 100 nanometres deep. On a perfectly ordered square array, mesenchymal stem cells showed little osteogenic response. When each pit was displaced randomly by up to 50 nanometres from its lattice position, the cells expressed bone markers and produced mineralised nodules. They did this without the dexamethasone and other supplements normally needed to drive bone differentiation.

Later work showed that the perfectly ordered array had its own effect: it helped stem cells retain multipotency over extended culture. This addressed a long-standing problem, since mesenchymal stem cells tend to drift or lose potency on flat plastic. A modest change in order, just tens of nanometres of offset, separated maintenance from differentiation. Similar size dependence appears elsewhere. On titanium dioxide nanotube arrays, tubes of about 30 nanometres supported adhesion without strong differentiation, while tubes of 70 to 100 nanometres induced cell elongation and osteogenic gene expression.

The explanation runs through the mechanotransduction chain described earlier. Topography that permits larger, more mature adhesions raises cytoskeletal tension, which favours osteogenic programmes via RhoA/ROCK signalling and nuclear YAP/TAZ. Topography that limits adhesion growth keeps tension lower, which favours fat-forming fates or quiescence. The cell combines adhesion size, adhesion distribution, and contractile state into a single fate decision.

The practical implications are significant. Growth factors are expensive, degrade in the body, and can act on tissues beyond their target. A topographical cue is built into the material itself. It persists as long as the surface does and acts only on cells in direct contact with it. For orthopaedic and dental implants, this means surface texture can be engineered to encourage bone integration directly. The idea now informs the development of nanostructured titanium implant surfaces.

Takeaway

Some biological instructions do not need to be molecules. A precisely placed pattern of pits can tell a stem cell what to become as clearly as a growth factor can.

Across adhesion, migration, and differentiation, a single principle recurs. Cells measure their surroundings mechanically, and nanoscale geometry sets the terms of that measurement. Ligand spacing limits clustering, clustering limits force, and force shapes behaviour.

This positions topography as a design parameter alongside chemistry and stiffness. The open challenges lie in manufacturing: transferring precise patterns from lithographically defined test chips to curved implants and three-dimensional scaffolds at scale. Another challenge is mapping how these cues combine when real tissues present all of them at once.

Feynman imagined arranging matter atom by atom. Biomaterials suggest a nearer goal: arranging it nanometre by nanometre, so that the surface itself carries instructions cells already know how to read.