Immunofluorescence (IF) is a milestone method which was introduced by Albert Coons in 1942. It uses fluorescent antibodies for tagging specific antigens in biological tissues. With this simple molecular complex, cell biology has been transformed into a visual science that borders on art.
The form, prevalent now in diagnostic laboratories, used to visualize antigens in fixed tissues has been termed immunohistochemistry (IHC). Introduced by Paul Nakane in 1967, it consists of a sequence of antibodies linked to enzymes that precipitate a chemical substrate. This precipitate is visible on tissues through a simple bright-field microscope, a device pervasive in most laboratories worldwide. Although the method has seen several iterations since, the fundamental tenets have held up for half a century.
So, what is needed?
IHC has been used consistently in the last 20 years to identify and subtype patient tumor tissues for diagnosis and treatment selection. With the dawn of personalized medicine and the growth of immunotherapy over the last few years, IHC has taken a more central role, as treatments are directly linked to antigens expressed by the tumors.
Considering its critical nature, there are still a few major limitations to IHC. The end-point of this assay includes signal amplification through enzyme-linked precipitation of a colored reagent. This specific step makes it very difficult to absolutely quantify the number of antigens in the tissue, as they are always obtained after reaction saturation, leading to visual signal variability and thus uncertainty in the diagnostic interpretation. Thus far, pathologists have subjectively scored IHC stains qualitatively and this has been sufficient. For the next generation of pathology, however, a more quantitative approach is needed and this is where IBM’s technology, the microfluidic probe (MFP) can play a critical role.
What is our contribution?
Using the MFP we perform microscale IHC on tissue sections with a gradient of incubation times for the primary antibody, and in doing so we achieve two significant advantages over the state-of-the-art including:
Pathologists always look at end results; in contrast we are able to perform reactions for a short time and track its kinetics to saturation in adjacent areas on a tissue section.
By only performing one critical step using the MFP, the rest of the workflow currently used for IHC remains unaltered, thus facilitating the integration of our method into current infrastructure.
We term this method quantitative microIHC (qµIC). The metric obtained through this method – a saturation approach matrix – characterizes the kinetic curve of the antigen-antibody reaction.
To improve the adoption of this methodology into current workflows:
We provide an easy-to-interpret visual representation of the kinetic curve. It can aid pathologists in differentiating different tumor samples and scoring them for biomarker expression when compared with pre-defined references, reducing uncertainty and increasing robustness and reproducibility.
We also automated the process from image acquisition to analysis, using supervised machine learning to output quantitative estimates of antigen expression when given qµIC images as input.
Through this work, we hope to stimulate the community to overcome bottlenecks in quantitative pathology while still being able to make use of current clinical workflows.
Team behind the work (L to R) Govind Kaigala (IBM), Maria Gabrani (IBM), Peter Schraml (USZ), Anna Fomitcheva Khartchenko (IBM), Aditya Kashyap (IBM) and Pushpak Pati (IBM).
Kashyap, A.; Fomitcheva Khartchenko, A.; Pati, P; Gabrani, M; Schraml, P; and Kaigala, G.V.Quantitative microimmunohistochemistry for the grading of immunostains on tumour tissues. Nat. Biomed. Eng. (2019). doi:10.1038/s41551-019-0386-3
At the 18th European Conference on Computational Biology and the 27th Conference on Intelligent Systems for Molecular Biology, IBM will present significant, novel research that led to the implementation of three machine learning solutions aimed at accelerating and guiding cancer research.
Our team of IBM researchers published research in Radiology around a new AI model that can predict the development of malignant breast cancer in patients within the year, at rates comparable to human radiologists.