Spatial Transcriptomics: Fixation and Permeabilization Trade-offs

Team Dynamic Matrices | 27 May 2026

Spatial transcriptomics asks a tissue section to do two things that pull in opposite directions: hold its architecture perfectly still, and let probes or capture oligos reach every transcript inside it. Fixation delivers the first and works against the second; permeabilisation delivers the second and works against the first. There is no setting that maximises both. There is only the right compromise for a given sample, and it has to be found empirically. This post is about how.

Why the failures are subtle

A badly optimised spatial protocol rarely produces an obvious error. It produces a plausible dataset with quiet problems: lower transcript counts in one region than another, a slight morphological distortion that throws off image registration, or a spatial expression pattern that is partly reagent penetration rather than biology. Under-permeabilised regions drop low-abundance transcripts first, so the map looks fine and the interesting genes are missing.

Fixation sets the terms

Aldehyde fixation (formaldehyde, PFA) creates methylene crosslinks between proteins and nucleic acids. More fixation (longer, higher concentration) means better morphology and denser crosslinking, which slows the diffusion of permeabilisation enzymes and probes and can mask target sequences. Methanol fixation avoids crosslinks and gives better RNA access at the cost of lipid extraction and some morphological change; it is the default for several array-based platforms for that reason.

For thin cryosections (10 µm) the fixation window is forgiving. For thicker sections, and especially for whole organoids or 3D cultures, diffusion distance grows, the periphery fixes before the core, and a protocol validated on sections will under-process the interior. Section thickness is therefore not an independent parameter; it is coupled to fixation and permeabilisation and has to be optimised with them.

Permeabilisation is the step to optimise

Because permeabilisation efficiency depends on fixation, and fixation depends on tissue, the permeabilisation step needs a time course for every new sample type. The standard approach, formalised in 10x Genomics' Visium tissue optimisation workflow, is:

1. Cut serial sections from the same block.
2. Permeabilise each for a different duration (for example 3, 6, 12, 18, 24, 30 minutes).
3. Measure two things on each: transcript capture (fluorescent cDNA footprint on an array platform, or housekeeping-gene probe signal on an in situ platform) and morphological integrity (H&E or DAPI, scored for nuclear sharpness and tissue continuity).
4. Pick the shortest duration that gives near-maximal capture without visible morphological loss.

Test both shorter and longer than the published recommendation for a comparable tissue. The optimum is usually where capture plateaus and just before morphology starts to blur.

Under-permeabilisation is the common failure

Researchers conservatively favour morphology when uncertain, so they stop short. The dataset still looks good. But capture efficiency is uneven across regions, and low-abundance transcripts drop out disproportionately, which distorts the apparent expression landscape and inflates the apparent dominance of highly expressed genes. This is very hard to detect without a positive control.

Organoids and 3D cultures need their own optimisation

Organoids in a hydrogel present three complications:

• Diffusion distance. A 300 µm organoid embedded in gel has a fixative and reagent path far longer than a 10 µm section. Fix longer, or section the embedded organoid before permeabilisation.
• The matrix. A dense protein gel around the organoid slows every reagent and can autofluoresce. Removing the matrix before fixation, or fixing and then dissolving it, improves access. DNA-crosslinked gels can be dissolved with a nuclease after fixation without disturbing the fixed tissue, which simplifies this.
• Embedding and sectioning. Organoids are small and easily lost. Pre-embedding in agarose or a defined gel before cryo-embedding keeps them oriented and sectionable.

Run the permeabilisation time course on organoid sections specifically. Do not transfer parameters from a tissue section.

Validate with a positive control before the real run

A housekeeping gene panel (ACTB, GAPDH, RPL genes) or a spike-in with known expected signal, run on one section under the chosen conditions, confirms adequate capture quantitatively before you commit an expensive slide set. If housekeeping signal varies by region more than tissue composition explains, permeabilisation is uneven.

Treat the protocol as sample-specific

The trade-off cannot be removed. It can be optimised for a sample type, and it must be re-established whenever the sample type changes: a new tissue, a new fixation, a new thickness, a shift from section to organoid. Budget a tissue-optimisation slide for each.