Journal Logo


Milestones in Immunohistochemistry and Molecular Morphology

Taylor, Clive R. MD, DPhil

Author Information
Applied Immunohistochemistry & Molecular Morphology: February 2020 - Volume 28 - Issue 2 - p 83-94
doi: 10.1097/PAI.0000000000000833
  • Free

Today, almost half a century from the first use of immunohistochemistry (IHC) in routinely processed tissues the method must be counted a success, beyond expectations. IHC is part of Surgical Pathology everywhere; essential in clinical diagnosis, pervading every surgical pathology textbook, intrinsic to much research and a key method in the great majority of published manuscripts in the field. Now is a good time to consider how the foundations of IHC impact the current practice of IHC, contributing to its strengths and its ongoing weaknesses, as an approach to moving forward to the higher levels of performance required for Precision Pathology. Just as the present finds its roots in the past, so also does the future.


Immunohistochemistry (IHC), as the name implies, seeks a constructive blend of immunologic and tissue-based methods. In its broadest connotation, IHC includes not only the more recent (40 y old) light microscopy “staining” methods but also immunofluorescence (IF) as the prototypic method (80 y old). This editorial highlights selected key discoveries, or “Milestones,” that were crucial to the evolution of IHC. They are Milestones (Table 1) encountered on my personal passage through the realms of Pathology and IHC, and during my stewardship of the Journal; others among you, on your parallel journeys, will have encountered some of these same milestones and others that differ. In retrospect it is remarkable that many of these milestones originated outside of the realm of Pathology; we pathologists evolved by the remarkable talents; to adopt and adapt.

Milestones in Immunohistochemistry

This editorial follows an invited presentation, “Immunohistology: A Historical Perspective,” given at the European Congress of Pathology in Nice, France in 20191 and draws at length from the multi-author text “Magic to Molecules: An Illustrated History of Disease.”3

Milestone 1: The Microscope “Invents” the Pathologist

The microscope had a prolonged gestation before achieving practical utility in biology and medicine.3–5 Galileo is reported to have observed, “flies as big as lambs” (in 1614). Robert Hooke was commissioned by Sir Christopher Wren to prepare microscope slides for King Charles II and depicted “cells” in cork in his book “Micrographia” (in 1665).3 Yet a further century passed with little advance, up to the time that John Hunter, surgeon and morbid anatomist, used a microscope in his studies of chick embryos. He was, however, dismissive of its value and of those who purported to have expertise in its use6—“Malpighi was probably the first who employed a microscope for this purpose (viewing the red globules of the blood) … in 1668 … in the blood vessels of the omentum, which he mistook, however, for globules of fat.” … “Microscopical observations were pursued with great ardour by Antonius Van Leeuwenhoek, who saw the red globules, August the 15th, 1673.” … “These early observers probably imagined more than they saw.6

Then in 1827, a noteworthy paper appeared, “Notice of Some Microscopic Observations of the Blood and Animal Tissues,”7 co-authored by Thomas Hodgkin and Joseph Lister. The former was to achieve great renown as a physician, the latter, was a wine merchant and amateur naturalist. Yet it was the wine merchant who arguably contributed more to the “invention of Surgical Pathology” (Fig. 1), for 5 years later Hodgkin did not use his microscope in his paper “On Some Morbid Appearances of the Absorbent Glands and Spleen, which described “His Disease,” whereas Lister went on the publish “On the Improvement of Achromatic Compound Microscopes,”8 which effectively reduced chromatic aberration, dispelling the globule theory of tissue structure. The quality of the image improved dramatically and adoption was then rapid, as evidenced by the foundation of the Royal Microscopical Society (1838), and the first course in histology at a medical school (Edinburgh, John Hughes Bennett, 1842). Some of the first recognizable histopathology texts (Sir James Paget 1854, Rudolf Virchow, 1858) (Fig. 2) appeared in short order …. “Thus the first pathologists emerged from the treacherous swamps of medieval practice onto the relatively firm ground that histopathology seemed to offer with respect to the diagnosis of disease.”4

Thomas Hodgkin Microscope by J. Smith of London, based on Lister design (circa 1840). From Accessed November 2019. Courtesy Dr Barry Sobel.
Two of the first anatomic surgical pathologists that were “invented”: Sir James Paget and Rudolf Virchow. By Spy (Sir Leslie Ward—from Vanity Fair) (1876 Paget; personal copy, Dr Amber Waits Taylor: 1893 Virchow; personal copy, Dr Emma Taylor).

Joseph Lister’s nephew, Richard Beck, was a leader in the commercialization of microscope production, resulting in widespread availability of affordable, good quality instruments (Fig. 1).3 The microscope had, at last, met the criteria necessary for practical utility,5 and as medical practitioners increasingly devoted time and energy to the device, so was the Pathologist “invented.” The reasons offered by Majno and Joris5 for this long gestation are relevant because they mirror almost exactly the “slogress” (slow progress) seen today in acceptance of digital pathology and the new “intelligent microscope” (Milestone 12) … “secrecy of the art, high cost of the apparatus, technical difficulty, the notion that the microscope was a toy, lack of new ideas, neglect by the Universities.”5

Milestone 2: The Clothing Industry Provides the Stains

The sudden enthusiasm to impart color to tissue slices for microscopy fortuitously coincided with the birth of the synthetic dye industry. Although it had long been possible to dye a lady’s gown blue, a King’s mantle red, or an Emperor’s toga Tyrian purple, using biological extracts from plants (hematoxylin, saffron) or animals (shellfish, cochineal carmine), the colors faded and many dyes were expensive, beyond the reach of the population at large. In the 1850s the aniline dye industry extended enormously the range of dyes, many of which were rapidly adopted and adapted as tissue stains.3 Exploitation of this borrowed technology was an outstanding success, evidenced by the hematoxylin and eosin (H&E) stain, attributed to an English Schoolmaster W.H. Poole in 1875,9 who combined hematoxylin from the logwood tree (Haematoxylum campechianum) with synthetic eosin (tetrabromofluorescein), and by many other stains from this era still in use today—Gram and Ziehl-Neelsen methods, Congo red, methyl violet ….

Alcohol served as the usual tissue preserving agent, with a long and distinguished history; the Irish ship’s surgeon, William Beatty, famously selected a cask of brandy to preserve Nelson’s body for the long return to London after his death at Trafalgar in 1805. Then industry provided yet another critical contribution to pathology when a saturated solution of formaldehyde gas (40%) became available in the late 1880s. Blum,10 experimenting with its properties as an antiseptic, found that a further tenfold dilution hardened the skin of his fingers, much as did alcohol, prompting a study of its use to “fix” tissues. Blum’s 4% formaldehyde soon became established in the “routine” of formalin-fixed paraffin-embedded (FFPE) tissue, the basis of surgical pathology ever since; a legacy for good and ill as far as IHC is concerned (Milestones 5 and 8).

Milestone 3: Labeled Antibodies—IF

By the turn of the century,1900, an H&E stain of an FFPE tissue section, examined by an experienced histopathologist, had become the “Gold Standard” for the diagnosis of many diseases, prominent among which was cancer.3 But an H&E stain has no intrinsic specificity for cell type, identification being subjective, dependent upon the experienced eye of the pathologist. Efforts were made to improve the range and cellular specificity of available “stains.” Histochemical (cytochemical) approaches sought to impart differential color to cells on the basis of active cytoplasmic enzymes (eg, acid phosphatase, nonspecific esterase), with success limited by difficulties in methodology and the fact that most intrinsic enzymes were inactivated by formalin fixation.

In 1938, the Medical Research Council published J.R. Marrack’s remarkable report, “The Chemistry of Antigens and Antibodies.”11 The field was ripe for invention and just 1 year later, Albert Hewett Coons, then a medicine resident at Massachusetts General Hospital made a seminal contribution. Coons found himself for a brief while at Charite Krankenhaus in Berlin, where he pursued an interest in the possible role of Streptococcal antigens in the formation of the Aschoff nodule of rheumatic fever. Famously he wrote—“It struck me that this theory had never been tested and indeed could not be tested without the demonstration of antibody or antigen, preferably both, in the local lesions. and … “The notion of labeling an antibody molecule with a visible label was perfectly obvious in this context.12

His paper, “Immunological properties of an antibody containing a fluorescent group,”13 served to launch the new research fields of autoimmune disease and immunopathology, based upon numerous studies using fluorescein-labeled antibodies. Yet decades passed and the IF method had a little direct impact upon diagnostic surgical pathology, notably excepting its use in renal and skin disease for the detection of immune deposits. The reasons were intrinsic to the method. Frozen sections were deemed necessary for IF studies and examination required a special microscope, filters and darkfield illumination. The net result was that the fine morphologic features established over 100 years of surgical pathology were rendered invisible; the “Gold Standard” upon which diagnostic pathology had been built was lost, and with it was lost the enthusiasm of the diagnostic surgical pathologist (Fig. 3A).

Oxford laboratory in 1974, staining for plasma cells. A, Frozen section, immunofluorescence, plasma cells appear green. B, Formalin-fixed paraffin-embedded section, immunoperoxidase with hematoxylin and eosin counterstain, plasma cells appear brown. C, Formalin fixed paraffin embedded section, immunostain (alpha naphthyl pyronine) with weak hematoxylin counterstain, plasma cells appear red.

Milestone 4: Peroxidase Conjugates—Immunoperoxidase and Light Microscopy

The eventual compatibility of the labeled antibody method with the morphologic “Gold Standard” was achieved by combining 2 changes; replacing fluorescein as the label and replacing the frozen section.

The attachment of an enzyme (eg, horseradish peroxidase, acid phosphatase) to the antibody as the label, in lieu of a fluorescent compound, provided a signal visible by electron microscopy and more importantly in the long run by orthodox light microscopy. The pioneering work in 1967 by Nakane and Pierce,14 was a major topic of the 1972 Gordon Conference on immune-electron microscopy. An independent and highly significant advance using similar labeled antibody principles led to the enzyme-linked immunosorbent assay (ELISA) method, developed by Stratis Avrameas, who attended that same Gordon Conference. The ELISA method remains today as a “Gold Standard” for measuring proteins in serum and fluids, a fact that carries a long-neglected lesson; namely that the immunoperoxidase method uses the same principles and reagents as ELISA, and in theory is perfectly suited for use in quantification, given that other test parameters are appropriately modified …. But this lamentation anticipates Milestone 10, encountered later along the road.

Milestone 5: The “Routine” FFPE Section

The adaptation of the horseradish peroxidase method to use in routinely processed FFPE tissues was the first milestone that I encountered personally in real-time. It was in retrospect crucial for 2 main reasons. First, there was the potential to retain the fine morphologic features upon which diagnostic surgical pathology had been based for greater than one hundred years; namely to preserve the “Gold Standard,” which would render the “immunoperoxidase” method attractive for use in diagnostic surgical pathology. Second, and of critical practical importance, the types of tissue generally available to pathologists when the need for special stains, including IHC, arose were, in fact, FFPE tissue blocks; thus when the stain was needed, it could be readily carried out.

The discovery that labeled antibody methods could be successfully applied to FFPE tissue was more a matter of “unlearning” previous misconceptions than of new learning. I was developing murine models of Hodgkin disease as part of a PhD thesis under the tutelage of Alistair H.T. Robb-Smith in Oxford, with the invaluable aid of Ian Burns, shortly to be joined by David York Mason (1941-2008). Identification of the abnormal affected cells in the murine model emerged as a critical issue. H&E morphology yielded only conflicting opinions (reticulum cells, histiocytes, others) among senior pathologists, which, if nothing else, showed that our model was perfect for the study of human Hodgkin disease, where exactly the same problem existed! Use of anti-immunoglobulin and antileucocyte antisera by IF methods in frozen sections gave positive labeling of some dark shadowy cells, the examination of which elicited a uniform response from senior pathologists in the department—“It’s no good old boy, I can’t see the cells” (Fig. 3A). Replacement of the fluorescent label with peroxidase yielded visible results in frozen sections, but morphology still was inadequate. In the face of overwhelming literature that FFPE precluded the use of labeled antibody methods, a series of experiments with different methods of fixation and paraffin embedment began. Cold alcohol fixation (Guy Sainte-Marie, 1962) was considered to hold the greatest promise. Routine FFPE was employed as a morphologic control and yielded the best results! (Figs. 3B, C).15–17 We had tripped over Milestone 5!

In attempting to extend this approach to other antigens in FFPE tissues, several obstacles emerged: the low sensitivity of the detection method, the limited range of antisera (antibodies) available, the adverse effects of formalin on many antigens, and the emotional attachment of established surgical pathologists to their hard-won H&E “Gold Standard.” The resolution of these problems provided the next 4 milestones.

Milestone 6: High Sensitivity Detection Methods; Peroxidase Antiperoxidase (PAP), Avidin Biotin Conjugate (ABC), Polymers

Directly conjugated antibodies gave poor results on FFPE tissues, and the indirect method provided only marginal improvement. Adoption of the elegant PAP method, described by Ludwig Sternberger at the Edgewood Arsenal in Rochester, NY, greatly improved detection sensitivity.18 Subsequently, PAP competed with the ABC method as to which had the greater ability to detect a broad range of antigens with greater intensity of the signal. Eventually, ABC became established as the more practical system, until beginning in the 1990s it was superseded by polymer-based labeling, which avoided problems of endogenous binding of biotin, while offering greater stability and higher sensitivity, and is now the standard method.

Milestone 7: Molecular Biology and Monoclonal Antibodies

In the early days of IHC, only a few antibodies (ie, antisera or polyclonal antibodies, or course) gave good results. These antisera were hard to prepare, purify and reproduce. The breakthrough came unheralded and once again found origin outside of traditional pathology, from a cluster of research laboratories in the Department of Biochemistry at the University of Cambridge.3 First, Fred Sanger identified the sequence of a protein (insulin), so providing a bridge to the conceiving of a structural model of DNA by Crick and Watson, leading back to Sanger again, in developing a method for determining the sequence of DNA (Sanger sequencing). In the same laboratories César Milstein and a visiting scholar, Georges Jean Franz Köhler, were working to understand how DNA and RNA translated to protein. Their approach was to create hybrids of cells containing the DNA information for producing a protein (lymphocytes; immunoglobulin) with a cell line that had the necessary machinery in place for producing the protein (murine plasmacytoma cells).19

These discoveries reaped a rich harvest of Nobel Prizes and in a little little more than two decades the face of biological research and medicine was changed forever.3 Molecular biology arrived and with it came monoclonal antibodies. Their uses are manifold, but in the context of this editorial IHC was given huge impetus. Not only was the range of antibodies greatly expanded (approaching infinity!), but it became possible to tailor the generation of monoclonal antibodies for use in IHC on FFPE. The potential was apparent to many. David Mason in the laboratory at Oxford was among the first to exploit it for IHC. Although another investigator, Alan Epstein, then at Stanford, launched a career-making literally dozens of monoclonal antibodies, moving to the University of Southern California (USC) where 1 project was to integrate IHC and monoclonal antibodies into the multiparameter lymphoma studies of Robert Lukes and John Parker.20,21

Milestone 8: Antigen Retrieval (AR)

Meanwhile, attempts were made to address the third obstacle to the more general use of IHC methods on FFPE tissues, by utilizing enzyme digestion to “unmask” or recover the antigenicity of proteins adversely impacted by fixation and processing. This approach showed improvement for a limited range of antigens, but methods were notoriously temperamental and difficult to reproduce. Shan-Rong Shi approached the problem from an entirely different direction that was a lesson to us all. In the late 1980s, he conducted an exhaustive review of the literature, the old hard way, closeted in dusty libraries. He uncovered work by Fraenkel-Conrat and colleagues in the 1940s describing the restoration of immunogenicity to tetanus toxoid, the toxin had been biologically “inactivated” by formalin as part of the process of making tetanus antitoxin for use in the treatment of war wounds. Often formalin treatment of the toxoid was excessive, compromising its use as an immunogen. The method to restore immunogenicity was … to boil it! (review by Shi et al22). Dr Shi discussed his findings with me at a meeting; I recall that my reaction was skeptical, that boiling would almost certainly denature the protein. Once again it was a matter of unlearning what we thought we knew (rather “what I thought I knew!”). We agreed that the proper recourse was for Dr Shi to perform the experiment. Boiling the FFPE section in buffer became AR, which almost magically expanded the range of antigen targets that could be demonstrated in FFPE sections (Fig. 4). AR spawned its own expansive literature and also had an unforeseen but far-reaching effect of contributing to the overall growth of molecular medicine, in that AR provides the basis for most methods of extracting protein, RNA and DNA for study by modern molecular methods.24

University of Southern California laboratory in 1991 showing one of the first anaplastic tumor cases studied using antigen retrieval. A, The initial stain for keratin, judged “negative.” B, The identical protocol for keratin plus antigen retrieval, judged positive, changing the diagnosis; shades of Gatter et al.23

Milestone 9: Value in Surgical Pathology Diagnosis

The potential for the development of a range of new “specific special stains” that promised a more objective means of cell recognition was apparent from the very beginning of the use of IHC on FFPE tissues.3,4,25 Progress was slow and there was resistance among senior established pathologists, who were comfortable in their morphologic expertise. These expert surgical pathologists followed their usual practice in examining an H&E section and “called them (the lesions) as they saw them.” So great was their reputation and credibility, that lacking any independent method for cell identification and diagnosis, their opinion, “what they called them” in an H&E section, was in practice synonymous with the definitive diagnosis; it was the “Gold Standard.” The single publication, in Lancet (Fig. 5),23 that undermined this long-held belief emanated from Kevin Gatter, previously one of our students in Oxford, working with David Mason. Gatter collected 120 consecutive cases of anaplastic small cell tumors from the Oxford department files and, on a blinded basis, subjected the recuts to just 2 IHC stains, for keratin and leukocyte common antigen. The result was clear and dramatic. Remarkably he was sufficiently brave, or foolish, as to publish it!23 Almost half the diagnoses of record for these anaplastic tumors, as made by experienced pathologists using traditional morphologic criteria on H&E sections, were simply wrong!

Lancet, June 8, 198523; a key paper in establishing the value of immunohistochemistry in surgical pathology.

Milestone 10: Quality Concerns: The “Total Test,” Automation, Controls, AIMM

Dimly perceived, this milestone was approached with limited enthusiasm and has yet to be passed in a manner that meets new and evolving needs (see Milestones 11 and 12).

IHC has become a standard tool in surgical pathology diagnosis everywhere and is part and parcel of all the major surgical pathology texts. Yet the very success of IHC exacerbated the problem of poor reproducibility among laboratories, locally and worldwide. In addition, “in situ hybridization” (ISH) appeared as an analogous method, first with fluorescent labels (fluorescence ISH) and then with chromogenic labels compatible with light microscopy (chromogenic ISH). However, in performing these tests quality control lagged far behind the standards routinely applied in the clinical laboratory.

In 1990 the Biological Stain Commission working with the Food and Drug Administration (FDA) began a process to address quality issues in reagent production and qualification.26 Concurrently the errors inherent in the manual performance of complex IHC test protocols began to yield to the improved performance and control that is inherent in automation, pioneered by David Brigati,27 and which is now the standard for major IHC vendors and major laboratories.

Subsequently, the idea of the “Total Test” approach26 that is routinely followed in clinical laboratory analytic testing was trialed by the group in our tumor research laboratory at USC (Fig. 6). Gradually the “Total Test” concept penetrated into anatomic pathology laboratories, bringing with it a greater focus on a selection of the proper controls, the beginnings of true quality control, and the notion of “fit for purpose” testing. The birth of external Quality Control programs, providing external arbiters of satisfactory performance had enormous impact, exemplified by CAP (College of American Pathologists) in the United States and more comprehensive programs offered by NordiQC ( and UKNEQAS (, both of which provide detailed feedback on reagent and protocol choice to participating laboratories. Other similar programs (eg, Canadian CIQC) are springing up worldwide, evidence at last of attention to the notion that improved outcomes can only be derived from improved standards of practice.

The University of Southern California Cancer Research Laboratory group in the mid-90s. Dr Shan-Rong Shi (second left front row) is seated on the author’s left, and Dr Richard Cote, the laboratories co-director, on the author’s right.

AIMM has played a leading part in the above process, by elevating expectations and standards through the recent quality control series of publications (Table 2), collaborations with external QC programs, and sponsorship of the formation of ISIMM (International Society of Immunohistochemistry and Molecular Morphology), having as one of its principal goals improvement in the reagents and methods intrinsic to the IHC total test.

The Immunohistochemistry “Control Series” of Papers Sponsored by ISIMM for AIMM

Milestone 11: Biomarkers: Prognostic and Predictive—the Advent of “In Situ Proteomics”

Upon its introduction into routine pathology in 1974 IHC was viewed as just another “special” stain,4,25,26 imparting color to assist morphologic diagnosis by the pathologist, albeit offering new levels of cell or tissue specificity. The efforts of Jules Elias, Craig Allred, and others in the early 1990s introduced the notion of semiquantitative IHC scoring (in reality, estimating the percentage of positive cells) for estrogen receptor (ER) and progesterone receptor (PR) in breast cancer. That IHC on FFPE sections supplanted cytosol based biochemical assays for ER and PR in clinical care was a harbinger of events to come, but acceptance was gradual and hard-won.

Then on September 25, 1998 (Fig. 7), everything changed.28 The FDA delivered letters to Genentech and to Dako with concurrent approval for the sale of the drug Herceptin (monoclonal antibody to HER2) and the Companion Diagnostic test, HerceptTest (detecting HER2), ushering IHC into the age of the Biomarker. With HER2 testing the impact was instantaneous and unfortunate. Laboratories everywhere rushed to implement the test, inevitably tweaking reagents and protocols including AR, as was their established custom for their usual IHC “stains” to achieve a staining intensity that “pleased” the individual pathologist, at the expense of whatever comparability among laboratories that might have existed in the first place.28,29

The beginning in ernest of the Biomarker era for immunohistochemistry.

In retrospect, 3 decades of utilizing IHC as if it was a simple stain26 had created across anatomic pathology laboratories a legacy of nonstandard methods and practices that at best were qualitative, and fell far short of the new demands for biomarker quantification that extended beyond basic ER/PR scoring of an IHC stain (ie, counting or estimating percentages). In 2004 representatives of several US organizations, including the FDA, plus invited experts, met at NIST (National Institute of Standards and Technology; in long lamentation over the dismal reproducibility of HER2 (and other IHC) testing. There was blame aplenty, with principal emphasis upon non-standard specimen handling, the need for improved controls, and the subjective nature of scoring by pathologists. The next decade saw a gradual return to basics30 with a realization of the need to consider, control and standardize all elements of the “Total Test,” including fixation, reagents, protocol and scoring, in an effort to convert an IHC stain into a truly quantitative assay. In essence, this approach was to take the elements of the quantitative ELISA assay and applying them to a tissue section, yielding a quantitative on tissue assay, that we have termed “in situ proteomics.”28,29,31,32

This effort renewed the focus on quality control concerns, the topic of Milestone 10, discussed previously. Two new areas of emphasis emerged.

The first was the issue of test validation, not just validation of the reagents, but the Total Test result.28,29 For example, FDA approved tests for programmed death ligand-1 (PD-L1) specify the tumor type, the drug (vendor and programmed cell death-1 or PD-L1 antibody), the clone used for testing, the protocol (including fixation, AR, reagent concentration and type of staining platform) and the scoring system [patterns of stain and percentage threshold(s)]. It soon became apparent that in many laboratories FDA approved tests (designated as In Vitro Diagnostic Devices) often were substituted by Laboratory Developed Tests (LDTs). In FDA parlance, LDTs (or home brews) are developed by an individual laboratory for use in that laboratory only. The reasons for a laboratory opting for LDTs are manifold, including lack of the specified platform, or the near-impossible challenge for a laboratory to set up a different approved test for each of several different programmed cell death-1/PD-L1 drugs, from different vendors, that clinicians wish to use. Furthermore, data from external QC programs show that even when approved tests are employed, in many instances the protocols are “adjusted” to seek more intense staining or to reduce costs, such that the approved test also becomes, in reality, an LDT. Not surprisingly clinical results obtained in the face of such variable practices are difficult to compare. Harmonization of these diverse approaches are facets of Milestones 10 and 11 that have yet to be passed.

The second area of emphasis was that scoring of the same cases (the same IHC stained sections) by different pathologists often gave different results, most critically around the threshold value(s) that determine a positive or negative result, and prescribe therapy, or not. At the 2004 NIST meeting, referred to above, there was open discussion of the issue of “subjective scoring” by pathologists, during which the idea was voiced that computer-assisted scoring might lead to greater objectivity and better reproducibility. That discussion related primarily to HER2 scoring, which in retrospect now seems relatively simple when compared with the multiplicity of PD-L1 tests and scoring systems available in 2020. However, the process for introduction and approval of new IHC biomarker tests virtually ensures that these test still are locked into the practices enshrined in the original 1998 Her2 test, and therefore inevitably carry many of the same deficiencies, especially with respect to attempting to reproduce the highly complex scoring systems that have evolved (exemplified by PD-L1).28–30

Milestone 12: Digital Imaging and Computer Assistance

The NIST discussion reported above had an effect that extended well beyond the confines of the meeting or the individual participants. It likely contributed to the decision by CMS (Center for MediCare Services in the United States) to provide additional payment for an IHC biomarker score generated with use of a computer algorithm; a decision which had the effect (by design or not) of stimulating the development of algorithms for IHC scoring, by virtue of adding an economic incentive. As noted, the 2004 discussion related to HER2 scoring, which is much less challenging than PD-L1 scoring, for which a pathologist may be expected to read one or more of several different approved (and nonapproved, LDT) tests, that are complicated by the scoring of immune cell reactions as well as tumor cells. The result is a plethora of different and complex protocols that are difficult to score reproducibly by the human eye (Fig. 8), however, experienced; but help is at hand in the guise of rapidly evolving digital artificial intelligence (AI) driven analysis, if we can overcome the hurdles to adoption.

A slide from the European Society of Pathology presentation1 demonstrating the near impossibility of scoring programmed death ligand-1 (PD-L1) to a 5% threshold. The denominator (total cancer cells) in a single high power (×40) field could be as many as 600 (if confluent and averaging 20 µm diameter). These cannot really be counted by eye; in the diagram they are estimated to occupy half of the field, giving an estimated denominator of 300. The numerator is the number of positive cancer cells (excluding immune cells); if 15, the “test” is positive and the patient is treated, if 14, negative! Pathologists are expected to do this in 10 fields, themselves subject to selection bias. In the opinion of the author this is not possible (go ahead, count them!), so we guess as reproducibly as we can.

Not for the first time, Pathology has been greatly assisted by developments outside of the immediate field, in Radiology and in the computer industry at large. The fact that by the turn of the millennium Radiology had “gone digital” and that many institutions were installing PACs (Picture Acquisition and Communication) systems, provided infrastructure and knowhow, and a readymade solution to the problem of moving around digital pathology image files, which in terms of size are several times larger than a computed tomography scan. Concurrent improvements in computer hardware and software were exponential (Moore law 1965; Gordon Moore, co-founder of Intel) and greatly accelerated the work of a handful of pioneering pathologists and vendors, who were working quietly in the background on digital and telepathology. A decade ago with the launch of new generation smartphones, such as the iPhone and its sister device the iPad, it was possible to conceive of the PathPad (Fig. 9) with a multiplicity of “Apps” to assist pathologists in scoring IHC and other aspects of image analysis such as tumor cell recognition.

At the 2011 Pathology Learning Conference in Florida sponsored by Dr Yazijii, in a moment of whimsy the notion of a digital PathPad was introduced, displaying the slide image and envisaging multiple “Apps” to assist pathologists in diagnosis. Today whimsy has been surpassed by reality in almost all aspects.

Continuing advances in scanning technology facilitated the approval of whole slide imaging for primary diagnosis.33 This advance when coupled with software management and data processing within the “Cloud” also offers the realistic option of scoring complex IHC biomarkers (such as PD-L1) by using AI-based algorithms on whole slide imaging, thereby obviating errors inherent in selection of a limited number of fields of view in manual scoring methods, and so providing assistance to more standard scoring by pathologists (Fig. 10).34,35 The Camelyon 16 and 17 studies (detecting micrometastases)36 demonstrated the improvements in accuracy to be gained by combining analysis using an AI-generated algorithm with the opinion of an expert pathologist: not quite perfection, but the closest approach to date, and a clear sign of the inevitability of the advent of the “intelligent” microscope.

Whole slide imaging combined with artificial intelligence digital read out for programmed death ligand-1 (PD-L1) scoring. Courtesy OptraScan from; Figure 3 Taylor et al.34 In its initial iteration the algorithm at least matched the pathologists. How long before Pathology has a “Big Blue” moment, recapitulating the superiority of IBM’s Big Blue computer over Chess Champion Gary Kasparov?


Twelve milestones, yet it feels that this is not the end; rather, to paraphrase Churchill, it is the end of the beginning. The last 3 Milestones in particular, while encountered, have yet to be successfully negotiated, and are critical if IHC is to achieve its unique potential of integrating the wealth of information inherent in morphology, with a tissue-based quantitative assay, dubbed “in situ proteomics.”

Attaining the level of standardization and quantification of IHC necessary for “in situ proteomics” will require an incremental improvement in the types of controls used in IHC. The current “in house” controls that are generated from archival tissues are by definition unique to each separate institution. Although they may suffice for the usual IHC stains (Table 2), they are strictly not comparable; they differ by date, tissue source and exact fixation and when used up are replaced by a new piece of tissue (that is a new control!). For the future archival controls must be supplemented by some form of universal quantifiable standard as proposed by Bogen,37 or by standardized “faux tissues” based on the “histioid” model,38 coupled with quantifiable internal or intrinsic reference standards comparable to those used in extraction based molecular assays to compensate for adverse effects fixation and processing.31,32

Last but not least is the “intelligent microscope.” Already many aspects of our daily life are facilitated by AI-based technologies, often subliminal, exemplified by navigation on the road, or online as to where we should shop and for what, and by “Siri” and “Alexa,” the harbingers of sophisticated “assistants” that will become available (and, I believe, ultimately mandatory!) for use by pathologists in the near and long term. In the more narrow context of this editorial, it is inevitable that AI digital assistants will facilitate the reading and scoring IHC tests, as an essential part of converting the principle of qualitative IHC “stains” to develop a range of closely controlled strictly quantitative IHC (ELISA-like) tissue-based assays—“in situ proteomics.”28,29,31,32

The appearance of Nikon Coolscope (Nikon) at several pathology exhibits in 2004 attracted queues of pathologists, who admittedly were mostly youthful in outlook and in years. It was in effect the precursor of a new breed of the microscope, an AI-supported digital “intelligent microscope,” that carries with it the potential to “invent” (reinvent?) Pathology and Pathologists, recapitulating that original birth of the discipline almost 200 years ago. Current practitioners of anatomic or surgical pathology are advised to heed the lesson of the past, to abandon “the treacherous swamps of medieval practice (ie, 200 y of subjective opinion-based pathology) onto the relatively firm ground that histopathology (ie, the intelligent microscope) seems to offer with respect to the diagnosis of disease.”4,26,28 Better be ready, or retire!

In this light, the decision to retire from the editorial role at AIMM at such an exciting time is tempered by optimism that the Journal, Applied Immunohistochemistry and Molecular Morphology, will continue to play a key part in the evolution of the Next Generation Pathologist. The new leadership of AIMM carries that charge into the future.

Clive R. Taylor, MD

Department of Pathology, Keck School of Medicine

University of Southern California

Los Angeles, CA


In almost 25 years as Editor, I have incurred debts beyond measure to individuals beyond number, whose contributions to the growth of AIMM dwarf any personal efforts. The ongoing collegiality, input and close friendship of my co-editor, Jiang Gu, has been essential to the strategic design of the Journal and our intent to achieve a global reach that encompassed the best work wherever the source. The Editorial Board is large and its members highly active in shaping the Journal and assuring quality; every manuscript is reviewed by at least one Board member. My fellow co-founders of the sister (or brother) society, ISIMM (The International Society of Immunohistochemistry and Molecular Morphology) provided impetus and direction beyond value in elevating the cause of recalibration and accuracy IHC. Wolters Kluwer-Lippincott Williams and Wilkins have provided a dedicated, interactive support team, among whom Kevin Anders has proved a long-term advisor and valued friend as he has risen through the ranks at the publishers.

Then last, and first, there has been the support, effort, patience, and enthusiasm of Susan Taylor, the Assistant to the Editor, without whom the past 25 years of stewardship would certainly never have happened; our marriage survives it and we step down together!


1. Taylor CR. 31st European Congress of Pathology, Nice, France. Symposium: History of Pathology: Colours of Pathology. Wednesday, September 11, 2019. Link to Congress Presentation. sy-26-004-immunohistology: a historical perspective. ISIMM members may also access a pdf file. 2019. Available at: Accessed December 2019.
2. Milestones. Nature. 2009. Available at: Accessed November 2019.
3. Van Den Tweel JG, Gu J, Taylor CR. From Magic to Molecules: An Illustrated History of Disease. Peking, People Republic of China: Peking University Medical Press; 2016.
4. Taylor CR. Immunomicroscopy: A Diagnostic Tool for the Surgical Pathologist. Philadelphia, PA: W.B. Saunders; 1986.
5. Majno G, Joris I. The microscope in the history of pathology. With a note on the pathology of fat cells. Virchows Arch A Pathol Pathol Anat. 1973;360:273–286.
6. Hunter J. A Treatise on the Blood, Inflammation and Gun-shot Wounds. London, UK: J. Richardson for G Nicol, Pall-Mall, London; 1794.
7. Hodgkin T, Lister JJ. Notice of some microscopic observations of the blood and animal tissues. Philosophical Mag New Series. 1827;2:130–138.
8. Lister JJ. On the improvement of achromatic compound microscopes. Philos Trans R Soc Lond. 1830;120:187–200.
9. Jones ML, Gal AA. The Age of the Microscope and the “Invention of the Pathologist” Chapter 32 in Reference 2, From Magic to Molecules: An Illustrated History of Disease. Peking, People Republic of China: Peking University Medical Press; 2016.
10. Blum F. Der formaldehyd als hartungsmittel. Z Wiss Mikrosc. 1893;10:314.
11. Marrack JR. Chemistry of Antigens and Antibodies Medical Research Council. London, UK: H.M. Stationery Office, Special Report Series No. 230; 1938:194.
12. McDevitt HO. Albert Hewett Coons A Biographical Memoir. Washington, DC: National Academies Press; 1996.
13. Coons AH, Creech HJ, Jones RN. Immunological properties of an antibody containing a fluorescent group. Proc Soc Exp Biol Med. 1941;47:200–202.
14. Nakane PK, Pierce GBJ. Enzyme-labeled antibodies for the light and electron microscopic localization of tissue antigens. J Cell Biol. 1967;33:307–318.
15. Taylor CR, Burns J. The demonstration of plasma cells and other immunoglobulin- containing cells in formalin-fixed, paraffin-embedded tissues using peroxidase-labelled antibody. J Clin Pathol. 1974;27:14–20.
16. Taylor CR, Mason DY. The Immunohistological detection of intracellular immunoglobulin in formalin-paraffin sections from multiple myeloma and related conditions using the immunoperoxidase technique. Clin Exp Immunol. 1974;18:417–429.
17. Burns J, Hambridge M, Taylor CR. Intracellular immunoglobulins. A comparative study of three standard tissue processing methods using horseradish peroxidase and fluorochrome conjugates. J Clin Pathol. 1974;27:548–557.
18. Sternberger LA, Hardy PH Jr, Cuculis JJ, et al. The unlabeled antibody enzyme method of immunohistochemistry: preparation and properties of soluble antigen-antibody complex (horseradish peroxidase-antihorseradish peroxidase) and its use in identification of spirochetes. J Histochem Cytochem. 1970;18:315–333.
19. Köhler G, Milstein C. Continuous cultures of fused cells secreting antibody of predefined specificity. Nature. 1975;256:495–497.
20. Lukes RJ, Taylor CR, Parker JW. Multiparameter studies in malignant lymphoma based on studies in 1192 cases. Proceedings of the 13th International Cancer Congress. Prog Clin Biol Res. 1983;132E:203–213.
21. Epstein AL, Marder RJ, Winter JN, et al. Two new monoclonal antibodies, Lym-1 and Lym-2, Reactive with human B-lymphocytes and derived tumors, with immunodiagnostic and immunotherapeutic potential. Cancer Res. 1987;47:830–840.
22. Shi S-R, Cote RJ, Taylor CR. Antigen retrieval immunohisotchemstry: past, present, and future. J Histochem Cytochem. 1997;45:327–343.
23. Gatter KC, Alcock C, Heryet A, et al. Clinical importance of analyzing malignant tumours of uncertain origin with immunohistological techniques. Lancet. 1985;1:1302–1305.
24. Shi ZR, Shi Y, Taylor CR, et al. New dimensions of antigen retrieval technique: 28 years of development, practice and expansion. Appl Immunohistochem Mol Morphol. 2019;27:715–721.
25. Taylor CR, Kledzik G. Immunohistologic techniques in surgical pathology—a spectrum of “new” special stains. Hum Pathol. 1981;12:590–596.
26. Taylor CR. Perspectives in pathology: an exaltation of experts: concerted efforts in the standardization of immunohistochemistry. Hum Pathol. 1994;25:2–11.
27. Montone KT, Brigati DJ, Budgeon LR. Anatomic viral detection is automated: the application of a robotic molecular pathology system for the detection of DNA viruses in anatomic pathology substrates, using immunocytochemical and nucleic acid hybridization techniques. Yale J Biol Med. 1989;62:141–158.
28. Taylor CR. Predictive biomarkers and companion diagnostics. The future of immunohistochemistry—‘in situ proteomics’, or just a ‘stain’? Appl Immunohistochem Mol Morphol. 2014;22:555–561.
29. Taylor CR. Quantitative In Situ Proteomics; a proposed pathway for quantification of immunohistochemistry at the light-microscopic level. Cell Tissue Res. 2015;360:109–120.
30. Hewitt SM, Robinowitz M, Bogen SA, et al. Quality Assurance for Design Control and Implementation of Immunohistochemistry Assays; Approved Guidelines—Second Edition. Wayne, PA: Clinical and Laboratory Standards Institute; 2011:1–139.
31. Taylor CR. Quantifiable Internal Reference Standards for Immunohistochemistry; the measurement of quantity by weight. Appl Immunohistochem Mol Morphol. 2006;14:253–259.
32. Taylor CR, Levenson RM. Quantification of immunohistochemistry—issues concerning methods, utility and semiquantitative assessment. Histopathology. 2006;49:411–424.
33. Mukhopadhyay S, Feldman MD, Abels E, et al. Whole slide imaging versus microscopy for primary diagnosis in surgical pathology: a multicenter blinded randomized non-inferiority study of 1992 cases (pivotal study). Am J Surg Pathol. 2018;42:39–52.
34. Taylor CR, Jadhav AP, Gholap A, et al. A multi-institutional study to evaluate automated whole slide scoring of immunohistochemistry assessment of programmed death ligand 1 (PD-L1) expression in non-small cell lung cancer. Appl Immunohistochem Mol Morphol. 2019;27:263–269.
35. Taylor CR. Whole slide Imaging—issues for use in diagnostic pathology: ‘routine’; stains, immunohistochemistry and predictive markers. Biotech Histochem. 2014;89:419–423.
36. Van der Laak J. Camelyon challenges 16 and 17; 2019. Available at: Accessed November, 2019.
37. Bogen SA. A root cause analysis into the high error rate in clinical immunohistochemistry. Appl Immunohistochem Mol Morphol. 2019;27:329–338.
38. Kaur P, Ward B, Saha B, et al. Human breast cancer histioid: an in vitro 3D co-culture model that mimics breast tumor tissue. J Histochem Cytochem. 2011;59:1087–1100.
Copyright © 2020 Wolters Kluwer Health, Inc. All rights reserved.