The World (of IHC) Is Not Enough

Immunohistochemistry (IHC) detects PD-L1 as a companion diagnostic for pembrolizumab, a humanized monoclonal antibody used in cancer immunotherapy. The assay provides a semiquantitative score pathologists can use to determine the likelihood of treatment success using PD-1/PD-L1 inhibitors.

Recall that the original test for estrogen receptors in breast cancer was a quantitative radioimmunoassay (RIA). It required relatively large amounts of fresh or cryopreserved tissue. IHC replaced the RIA method largely because it fits into the existing histology workflow for formalin-fixed, paraffin-embedded tissue, making its convenience obvious. Histopathology became the go-to approach for personalizing breast cancer treatment, and soon, the HER2 assay also became part of the tissue pathology toolkit.

IHC offers a powerful way to determine whether a protein is present in cells or tissue – and where that protein of interest is located. But how much is present? We’re able to say to some extent, but accurate quantitation remains challenging. Morphometric algorithms and technical automation don’t overcome the intrinsic variables that affect IHC, such as fixation, tissue processing, antigen retrieval, antibody avidity, antibody titer, and chromagen development. So is it reasonable to assume that IHC can deliver everything we need?

Smaller molecules (<2000 kD), such as metabolites, are largely undetectable by IHC. Metabolites are part of the wider “omics” landscape, and metabolic changes have been associated with cancer since the initial description of the Warburg effect in 1956. Metabolomic studies have really evolved since those early days, and my colleagues and I have recently developed a way of incorporating metabolomics into the histopathology workflow by using alcohol as a primary tissue fixative, followed by secondary fixation in formalin and embedding in paraffin. The alcohol extracts metabolites and lipids, but preserves tissue architecture and allows proteins, RNA and DNA to remain in the tissue. Then, we use liquid chromatography-mass spectrometry (LC-MS) to determine the metabolites in the alcohol. Thus, we can perform repeated microscopic analyses on the same tissue – just as we have always done.

Our new method, for which we have coined the term “histabolomics,” overcomes several hurdles in the application of metabolomics to human tissues. One such hurdle is the small size of human tissue biopsies; an assay that competes for tissue will not be welcomed onto the playing field. Another is the need for normalization – the expression of the quantity of an analyte per unit of sample. In clinical chemistry, the analyte is expressed per volume of serum or plasma; tissue RNA is often expressed in relation to a housekeeping gene. But all of these approaches require extraction and disruption of the tissue. Normalization of metabolomics data is usually performed “post-acquisition,” when the LC-MS data are analyzed, processed, and normalized in relation to total ion counts or similar values – or by tissue weight. Even if the tissue is weighed, the amounts of disease or tumor relative to non-diseased tissue and stroma remain variable. Histabolomics downscales the method to accommodate as little as 5 mg of tissue, the typical yield of an 18-gauge core needle biopsy. This bypasses the need for tissue cryopreservation, commonly used for metabolomics. When combined with a chemical labeling technique, the method is quantitative and normalized.

Histabolomics is complementary to existing methods of RNA and DNA analysis, and to in situ methods such as FISH or CISH. Both tissue metabolomics analysis and routine histopathology and IHC can be performed on exactly the same tissue. Although it’s inconceivable to picture a future without IHC, it is entirely possible to imagine one where quantitative, normalized metabolomic data from human tissues is combined with histopathology, DNA sequencing, RNA expression and IHC to enhance clinical decision-making.

Such an approach is useful in distinguishing aggressive from indolent cancers, and my colleagues and I suggest that it could also be applied to other medical conditions where we operate with sparse or imperfect data – inflammatory bowel diseases, liver diseases, identification of drug targets, toxicology and more. In my opinion, pathologists should pursue diagnostic methods that yield as much information as possible, with as little impact as possible on the patient – and our histabolomics approach fits nicely into that picture.