Challenges with RNA Extraction from FFPE Samples

Team Dynamic Matrices | 28 July 2026

Formalin-fixed, paraffin-embedded tissue is the largest archive of human disease that exists. Hospital pathology departments hold decades of blocks with clinical follow-up attached. The same chemistry that preserved the morphology damaged the RNA, and the tension between those two facts shapes every FFPE sequencing workflow. Here is what the damage is, how to measure it, and how we suggest choosing a library strategy.

How formaldehyde damages RNA

Formaldehyde forms methylene bridges between amine groups on nucleic acid bases and on proteins. That locks tissue architecture in place, and it covalently traps RNA in a protein network. Three things then happen over the life of the block:

1. Fragmentation. Hydrolysis at elevated temperature during processing and slow degradation in storage reduce most transcripts to fragments of 100 to 300 nucleotides. Older blocks are worse.
2. Adducts and crosslinks that survive extraction. Monomethylol adducts on bases stall reverse transcriptase, giving truncated cDNA. Von Ahlfen et al. (PLOS ONE, 2007) showed how fixation time and storage govern this.
3. Loss of poly(A) tails. Fragmentation clips the 3' end from the body of the transcript, so poly(A) selection captures a small, 3'-biased fraction.

Fixation duration, block age, storage temperature and pre-fixation ischaemia time all shift the balance. Two blocks from the same tumour can differ substantially.

Why standard kits struggle

Column and bead kits were optimised for intact, abundant RNA. Applied to FFPE they hit three problems:

• Crosslink reversal is a trade-off. Heating in Tris buffer at 70 to 80 °C reverses methylene bridges but accelerates fragmentation. Too little heat leaves RNA trapped; too much shreds what is left.
• gDNA co-purifies. Crosslinked genomic DNA fragments overlap with RNA in size, so size-based cleanup does not separate them. A DNase step is mandatory, and it has to be effective on crosslinked DNA.
• Yield is low and variable. Recovery of 10 to 100 ng from a 10 µm section is common, and losses at every bind, wash and elute step take a larger fraction of a small input.

QC: DV200, not RIN

RIN was designed around intact 18S and 28S peaks and is meaningless for FFPE, where those peaks are gone. Illumina introduced DV200, the percentage of fragments longer than 200 nucleotides, and it correlates with library success far better (Illumina technical note, 2016; Matsubara et al., PLOS ONE, 2020). Our working bands:

Measure DV200 on every block before library prep. It costs minutes and it saves failed libraries.

Choosing a library strategy

Poly(A) selection is the wrong default for FFPE. The tails are clipped, coverage is confined to the 3' end, and yield collapses.

rRNA depletion keeps whatever fragments exist, regardless of whether they carry a tail, and is the standard choice down to about DV200 30 percent. The catch is that many depletion chemistries were validated on intact RNA. Probe-based methods lose efficiency on fragmented targets because a 100-nucleotide fragment may not span the probe site; RNase H methods leave short rRNA fragments that escape digestion. Residual rRNA then consumes reads you cannot afford. We built riboLASSO to capture rRNA fragments by sequence through polymer phase separation rather than enzymatic cutting, and it holds efficiency on degraded input where enzymatic methods drop; we cover the comparison in our rRNA removal method post.

Targeted capture flips the logic: instead of removing what you do not want and sequencing the remainder, you capture the transcripts you care about. For FFPE at low DV200 this concentrates depth on the fragments that matter. It requires that you know the question in advance, but for clinical panels and validated signatures that is usually true.

Practical recommendations

1. Record block age, fixation time if known, and section thickness. Match the crosslink-reversal step to block age where the kit allows.
2. Use a kit or protocol with a validated, extended DNase step and confirm gDNA removal by no-RT qPCR.
3. Measure DV200 before committing to library prep. Do not use RIN.
4. Default to rRNA depletion with a chemistry validated on fragmented RNA, or targeted capture. Avoid poly(A) selection.
5. Increase input where you can; 50 to 100 ng is a reasonable floor for whole-transcriptome work.
6. Run an FFPE control block of known quality alongside each batch to separate extraction failures from sample failures.