Fluorescence in situ hybridization grew out of the demonstration that nucleic acid probes carrying a hapten could be detected on chromosomes without radioactivity, and that such probes allowed quantitative analysis of interphase nuclei as well as metaphases (Pinkel et al., 1986). Its application to archival pathology material made it a routine tool for detecting gene amplification and rearrangement in solid tumors. Guidelines for HER2 testing in breast cancer and for ALK testing in lung cancer brought the method into almost every diagnostic laboratory (Wolff et al., 2018; Lindeman et al., 2013).

Many patients who need these tests never have a surgical specimen. Their diagnosis is made on a pleural fluid or an aspirate, and the material for ancillary testing is a cell block: a pellet of cells fixed, embedded in paraffin and sectioned like tissue. Guidance on thoracic small samples accepts cell blocks for molecular and in situ studies and stresses that their handling must be known and controlled (Roy-Chowdhuri et al., 2020). In practice a cell block is a less uniform substrate than a tissue block, and protocols written for tissue sections do not always transfer.

The commonest technical failure is a slide with weak or absent signals in nuclei that otherwise look intact, or bright signals in nuclei so damaged that they cannot be counted. Both are failures of probe penetration and its control. This article proposes a stepwise procedure by which a laboratory can optimize penetration for its own cell blocks. It describes a method and its rationale. No experimental results are presented, and the procedure has not been validated here.

Formaldehyde fixes by forming methylene bridges between reactive groups on proteins and between proteins and nucleic acids, and the reaction continues for many hours after the fixative has penetrated (Fox et al., 1985). The result is a dense network around the chromatin that preserves morphology and excludes large molecules. Prolonged fixation deepens the network and also damages nucleic acids (Srinivasan et al., 2002), while delay before fixation allows degradation that no pretreatment can repair (Hewitt et al., 2008). Labeled probe fragments, typically a few hundred nucleotides long, must cross this network to reach denatured target DNA. Figure 1 shows the barriers in the order in which a probe meets them and the step that deals with each.

Pretreatment reverses enough of the fixation to let the probe in. After dewaxing, sections are heated in a buffer or a chaotropic solution, a step that owes its logic to the heat-induced retrieval introduced for immunohistochemistry (Shi et al., 1991), and are then digested with a protease, usually pepsin at acid pH or proteinase K. Published protocols for paraffin sections and tissue microarrays agree that these two steps, and above all their duration, decide the outcome (Chin et al., 2003; Summersgill et al., 2008). Too little treatment leaves the target covered. Too much dissolves the nuclear proteins that hold chromatin in place, and the DNA is lost or the nucleus collapses.

Three features set cell blocks apart from tissue. The first is the variety of preparation methods. Cells may be clotted with plasma and thrombin, suspended in agar or a commercial gel, collected in a collodion bag or processed by an automated system, and each matrix has its own permeability (Jain et al., 2014; Krogerus and Kholová, 2018). The second is pre-fixation. Samples often reach the laboratory in an alcohol-based transport or liquid-based cytology medium and meet formalin only afterwards. Alcohols coagulate proteins instead of cross-linking them, so a block of this kind may need a different pretreatment from one fixed in formalin alone (Shidham, 2019).

The third feature is the absence of stroma. Cells in a block lie singly or in small groups, with little collagen around them, and are reached by reagents faster than cells deep in a tissue section. A digestion time that suits a core biopsy can over-digest a cell block. Cellularity is also uneven, and good blocks depend on concentrating the cells in one plane of section (Nathan et al., 2000). The consequence for the laboratory is that conditions must be established on its own blocks, made by its own method, and cannot be copied from a tissue protocol or from another institution.

The procedure has six stages. It assumes a laboratory that already performs the assay on tissue sections and wishes to extend it to cell blocks, or one whose cell block failure rate is unacceptable.

A laboratory with few cases can compress the scheme. The factorial grid can be reduced to two levels of each factor, and stage 5 can be met by accumulating cases over time, provided each new block type is run alongside a control. Laboratories using automated stainers have fewer free variables, and for them the procedure reduces to selecting among the pretreatment programs the instrument offers, scored by the same rubric. Where the assay is a regulated companion diagnostic, changes to the validated method may not be permitted, and the proper use of the procedure is then to define which cell blocks are acceptable for the unmodified assay. Pre-analytical control remains the most effective intervention of all: prompt fixation in neutral buffered formalin for a duration within the range recommended for the test (Wolff et al., 2018) removes much of the variation that optimization would otherwise have to absorb.

The procedure is a proposal, and its benefit over the informal trial and error that most laboratories use has not been measured. A fair test would compare failure rates and inter-observer agreement before and after its adoption across several laboratories that use different cell block methods. The rubric is ordinal and partly subjective; image analysis of signal-to-background ratio would be more objective but is not widely available. Control blocks made from cell lines or pooled effusions may not reproduce the behavior of scanty aspirates. And the scheme optimizes for one probe at a time, whereas probes differ in size and labeling, so a condition that suits a centromeric probe may be marginal for a small locus-specific one.

Probe penetration in formalin-fixed cell blocks is governed by a balance between unmasking the target and preserving the nucleus, and the balance point depends on how the block was made. The procedure proposed here asks a laboratory to characterize its own blocks, to judge slides by fixed written criteria, to tune heat pretreatment and digestion jointly, and to verify the outcome across the block types it receives before setting cut-offs. Each element is taken from established practice. Their combination into a defined sequence remains to be evaluated.

To be declared by the author(s) before publication.

To be declared by the author(s) before publication.