Flat Biology Is Lying to You: The Case for Leaving 2D Behind

Team Dynamic Matrices | 27 December 2025

A cell on a polystyrene dish is not resting in a neutral, simplified version of its natural state. It is responding to an environment that exists nowhere in the body: gigapascal stiff, perfectly flat, flooded with oxygen and nutrients, and missing most of the contacts that tell it what kind of cell to be. We all know this and we use 2D anyway, because it is cheap, fast and compatible with every plate reader built since the 1950s. What has changed is the evidence for what that convenience costs. Flat culture does not simplify biology; it distorts it, often in the direction that produces false confidence in a drug candidate.

The flattening problem is not cosmetic

When a cell adheres to plastic it spreads, polarises wrongly and loses the cell-cell and cell-matrix contacts that organise tissue. Genome-wide comparisons of the same cancer lines in 2D and 3D (Zschenker et al., PLOS ONE, 2012) show altered expression of genes governing integrin signalling, cell shape and cell-cell contact, and cytochrome P450 drug-metabolism genes are deregulated in spheroids relative to monolayers. If the drug-metabolising machinery differs between 2D and 3D, every pharmacokinetic inference from a flat screen rests on altered biochemistry, not a simplified version of the real thing.

Mechanics compounds it. Tissue culture plastic has a modulus around 1 GPa. Brain sits at about 0.1 to 1 kPa, liver a few kPa, muscle around 10 kPa (Engler et al., Cell, 2006). Since integrin mechanotransduction shapes differentiation, proliferation and drug sensitivity, plastic is not withholding a variable, it is supplying the wrong one. A cell on a dish is not experiencing "no mechanical signal". It is experiencing the mechanical signal of concrete.

Where the lie shows up in the data

Oncology gives the clearest evidence because the cost of getting drug response wrong is direct. Breslin and O'Driscoll (Oncotarget, 2016) showed HER2-positive breast cancer lines in 3D were resistant to trastuzumab, lapatinib and taxanes in ways their 2D counterparts did not reproduce. Jabs et al. (Molecular Systems Biology, 2017) found that in organotypic 3D cultures, oncogene addiction and EGFR-inhibitor efficacy tracked patient response, while 2D showed only an attenuated version of the same signal. Riedl et al. (Journal of Cell Science, 2017) documented that PI3K/Akt/mTOR signalling and drug response differed between 2D and 3D in colon cancer lines because of the altered mechanical and diffusion context.

Diffusion is the other thing 2D cannot represent. In a monolayer every cell has unrestricted access to drug, oxygen and nutrients. In a tumour or an organ, cells sit along gradients, and the cells furthest from a vessel see a different concentration, a different oxygen tension and often a different phenotype. A 2D assay cannot produce that gradient because there is no depth for it to form across. Screening efficacy without a diffusion barrier is screening a question the drug will never face.

Why the default persists

Partly infrastructure: decades of automation and imaging built around flat-bottom plates. Partly cost and throughput. Partly inertia: familiarity gets mistaken for validity, especially when the downstream failure, a phase 2 flop for reasons the 2D screen never flagged, is organisationally distant from the bench where the screen ran. The economic case for 2D is real, but it is a case about convenience, not accuracy.

The honest framing is not "3D is better". It is that 2D makes specific claims about mechanics, gradients and cell shape that are almost never true of the tissue being modelled. Every time a lab defaults to flat, it implicitly asserts those variables do not matter for the question. Sometimes that is a defensible bet. Often it is the bet nobody examined.

How we suggest making the switch

You do not need a full organoid program on day one. What we recommend to labs starting out:

1. Pick one assay you run often and repeat it in 3D alongside 2D for three weeks. Same cells, same drug, same endpoint. Seeing the IC50 shift in your own hands is worth more than any review.
2. Use a matrix you can pipette like a coating. The main historical barrier to 3D was handling: ice buckets, viscous gels, timed pipetting. Defined thermoresponsive matrices remove that. DyNAtrix, for example, is mixed with cells at room temperature and sets at 37 °C.
3. Keep the matrix defined so the 2D-to-3D comparison is about dimensionality, not about undefined growth factors.
4. Report the mechanics. Stiffness and, if applicable, stress relaxation. It makes the experiment interpretable and it makes reviewers happy.

The burden of proof should sit with the flat model. The tools to move past it already exist. What is missing is the willingness to stop treating flat as neutral.