From Stain to Shining Stain

For nearly 80 years, immunohistochemistry (IHC) has been a valuable tool for pathologists and laboratory medicine professionals the world over. In every stage of its evolution, we owe it many diagnostic debts – but I’m not here to document the complete history of IHC. For that, there are several excellent papers (1,2,3) to which I direct curious healthcare professionals. Instead, I would like to present a more personal experience: a journey I’ve shared with histotechnologists that traverses nearly five decades of IHC advances – from the early 1970s to the present day. After almost half a century, it feels appropriate to document the progress we have made thus far and note where we still have room to improve.

The IHC technique was developed from immunofluorescence (4,5,6) to light microscopy with the use of peroxidase/antiperoxidase (7) – an approach sometimes informally known as the “hamburger” because of the way antigen and antibody are sandwiched. My colleagues and I started using this method in the 1970s with our own, often very basic, problems.

Preanalytic parameters

Adherence of tissue sections to slides

To get tissue sections to adhere to the slides during IHC, we tried several adhesives: white glue, egg white, and serum – each a messy and unreliable solution. Positively charged slides were a markedly better proposition, although we had to dip each one individually into polylysine or silanize it with a solution of 3-aminopropyl triethoxysilane in the presence of catalytic traces of water (8). Later, positively charged slides became available commercially; now, they are cheap enough to be used routinely on all slides cut in the histology laboratory – and so we’re able to perform IHC over a hematoxylin and eosin (H&E)-stained or previously IHC-negative slide without de-staining or losing the section.

Optimization and validation

Originally, antibody optimization required the use of one or two positive control tissues in a checkerboard pattern with the other reagents just to ensure that the antibody worked. Nowadays, the process is more involved; we also perform validations (on 10 positive and 10 negative tissues wherever possible) to determine sensitivity and specificity. We have made a tissue microarray (TMA) of many different neoplasms and normal tissues that serves as our negative control and also shows tissues that are unexpectedly positive with the antibody.

There are now guidelines for the validation of all antibodies (non-predictive markers) prior to use on patient specimens (9), with at least 90 percent overall concordance required between the new diagnostic test and the comparator test or expected results. Users must test a minimum of 10 positive and 10 negative cases (10). For predictive markers such as estrogen and progesterone receptors (ER and PR) and HER2, even more validation is required – users should include 20 positive and 20 negative cases with ±95 percent concordance (11,12). Other quality parameters include proficiency testing, which we obtain from the College of American Pathologists (CAP) for ER, PR, HER2, c-KIT, microsatellite instability (MSI), and hematopoietic markers. Online databases compare antibodies from different vendors, published literature indicates antibody sensitivity and specificity (although many papers omit what positive control is used), and accreditation processes all help acquire and maintain quality IHC.

Conferences and publications

IHC was, and still is, incorporated into general and subspecialty meetings. I was there when the United States and Canadian Academy of Pathology (USCAP) conference first became a “brown” meeting because of the many posters that used diaminobenzidine (DAB) as a chromogen in IHC. Today, the move is more toward in situ hybridization (ISH) and molecular studies (13,14). As a result, events specifically focused on IHC are finding a place. Since 2007, Hadi Yaziji and Richard Eisen have organized an excellent annual retreat exclusively for technologists and pathologists using applied IHC and molecular pathology (15), and a meeting of the International Society of Immunohistochemistry and Molecular Morphology (ISIMM) is in the pipeline.

IHC has also long been included in both general and subspecialty journals. Publications specifically designed for IHC include Applied Immunohistochemistry and Molecular Morphology (AIMM), available online and as a hardcopy, and the forthcoming ISIMM Journal. For those who prefer to learn online, the Pathology Outlines website gives the expected immunoprofile for each neoplasm its authors discuss. Immunoquery gives different antibodies and combinations with expected sensitivity and pertinent references for various neoplasms. ExpertPath (expertpath.com) is an online alternative to maintaining multiple sets of hard copy reference materials.

Analytic parameters

Antigen (epitope) retrieval

Antigen retrieval to overcome the effects of formalin fixation has undergone an impressive evolution. Initially, we used a five to 10-minute trypsin incubation; later, we moved to heating via a hot plate, which improved IHC, but resulted in background stain. We improved our results further by switching from the hot plate to 20–30 minutes in a microwave, steamer, or pressure cooker for most IHC, or 30–40 minutes in a hot water bath for HER2 (16). Much like early polymerase chain reactions (PCR), we began by performing all of these heating steps manually – but modern technology has given us the automated stainer, which controls time, temperature, and pH.

Antibodies

Antibodies were difficult to obtain in the early days of IHC. The best were those homemade by basic researchers, which we could dilute to 1:10,000 or more. We did try antibodies used in radiology and bought several from commercial companies – but they were like water; they never worked. A commercial alpha-fetoprotein antibody, given to me in 1972 in Copenhagen, did not work on several fetal livers of different months’ gestation, despite what was written in the literature. The magic bullet was the alpha-1-antitrypsin antibody (AAT), which stained a liver from an AAT-deficient patient and set us on the path to success.

Since then, many antibodies have become available from many different and often very good companies. We even have ready-to-use antibodies for those who prefer them to concentrated antibodies requiring titration – and some are formulated especially for automated stainers. These make life easier for histotechnologists, who can now work more efficiently and with greater ease.

We have also improved antibody quality by refining our sources. From polyclonal rabbit anti-human antibodies, which resulted in a certain amount of nonspecific background, progress led us to mouse and then rabbit monoclonal antibodies, which are more sensitive and have less background (17).

IHC methods

We began with the direct immunolabeling technique, in which the labeled primary antibody detects tissue antigen with a one-step protocol (not as sensitive as modern protocols and with nonspecific background). The indirect method of immunolabeling followed; in this approach, the bivalent primary anti-human antibody binds to cell or tissue antigens and to the labeled secondary antibody, which reacts with IgG from the species of the primary antibody. The indirect immunoenzyme peroxidase technique is similar – but the secondary antibody has a peroxidase label attached that causes DAB substrate to precipitate over hydrogen peroxide when both are introduced. The immunoenzyme-bridge technique has a third antibody made in the same species as the first and directed against peroxidase; this method uses the sandwich approach, in which peroxidase is added, followed by the peroxidase substrate. And, finally, in the peroxidase-antiperoxidase complex (PAP) technique, the third antibody to peroxidase is added in a complex with peroxidase (7), which leads to increased sensitivity and decreased background. Ultimately, a long dextran polymer labeled with secondary antibody and multiple enzyme molecules can further increase sensitivity and avoid endogenous background biotin.

These techniques use similar enzymes: horseradish peroxidase, which is highly sensitive and gives a precise chromogenic reaction, and alkaline phosphatase (AP), which is useful for tissues with pigment (as in dermatopathology) and for double staining. Originally, the chromogen of choice was the brown DAB, which is alcohol resistant. It was eventually suspected to be carcinogenic and its removal from the market was planned; although this never happened, we nevertheless changed to amino-ethyl-carbazole – but this red stain is not alcohol-resistant, resulting in a loss of stain when we used Permount rather than Aquamount for coverslipping. Today, we use Fast Red TR, which is alcohol-resistant, for dermatopathology specimens.

Automation

Our IHC run was initially performed manually in one or two 10x13” cake tins bought for less than US$10 each at a department store. We placed a damp hand towel under the metal slide tray. This assembly could be incubated at room temperature for 30–60 minutes or overnight at -4°C. We sometimes used a slide agitator, which we hoped would result in evenly distributed stain on the slides (it did not). Once we needed to stain more than 40 slides (in two cake tins), we moved to automation. The technologist could not put one reagent on more than 40 slides and then return to add the next reagent without rushing or letting slides dry out, and mistakes inevitably occurred.

Elizabeth Unger worked with me in the anatomic pathology department at Emory University. She and the late David Brigati developed the first automated slide immunostainer, the Codon (18,19), which used capillary action. We then acquired a 250-slide stainer, which we nicknamed “Hartsfield Airport” for its size. Unfortunately, we found that it took hours to complete a run, so we only stained 100 slides at a time – spurring us to purchase a smaller 100-slide stainer for more practical use.

Around this time, we stopped being able to use microwave antigen retrieval, our standard technique prior to a run. Why? Because it was a patented technique... (Who allows people to patent a scientific method?!) As a result, we used first the steamer and later the pressure cooker. Fortunately, all of these heat antigen retrieval methods produced similar results. We performed our ER and PR IHC on formalin-fixed sections using the Codon (20,21). Results were not excellent, but they were interpretable (especially if positive), and no fresh frozen tissue was available. We also immunostained immunoglobulins and complement in fixed, paraffin-embedded sections of kidney biopsies. Results were good, but the technologist never instituted the new technique in the fluorescence laboratory, so we never got to see its performance on a larger scale.

Eventually, we got a 48-slide Dako autostainer that proved to be excellent. We currently have three and love them, although we do need to manually retrieve antigens beforehand. In addition, we use four Leica 30-slide stainers on which we can also perform antigen retrieval according to the necessary parameters – at low or high pH for 10 or 20 minutes. We use concentrated antibodies that we optimize ourselves wherever possible, but many excellent ready-to-use antibodies are also available. Finally, we have also used a Ventana Benchmark autostainer, which yielded excellent results for a research project on BRAF in melanomas (22). Each automated system has its pros and cons (23,24) – some are closed systems that incorporate antigen retrieval, but require company service for problems rather than being reparable in-house. Each laboratory must assess and choose according to its individual needs and budget.

Controls

For many years, each antibody required positive, negative, and internal controls (25,26). We have used tissues remaining from clinical surgical specimens, prior to incineration, for positive controls; we get blocks as soon as possible so as not to over-fix in formalin. If patient consent is required, as is now being suggested, there may be a problem for old blocks we still use, which only have a tissue and/or tumor type and sometimes a surgical pathology number indicated. Patient names are never used. Currently, for clinical research projects, patient consent is obtained by clinicians before any surgical procedure – which will make life easier for my colleagues down the line. We do occasionally get blocks of tonsil (from a pediatric laboratory) for the many hematopoeitic markers, syphilitic lesions (for Treponema pallidum), placenta (beta-hCG), and Kaposi’s sarcoma (HHV8) to name a few, and we give blocks to other IHC laboratories as well.

Today, we are testing synthetic tissues for use as positive controls (27). The synthetic tissues are in a small TMA from two sites and two tumors; each core stains well for numerous antigens and reproducibility over five runs, when quantitated by image analysis, is excellent. We are now testing a TMA containing synthetic HER2 controls with three different levels of HER2.

We have used small TMAs that we have made from tissue cores of breast or endometrial carcinoma with negative, low, intermediate, and high ER, PR, and HER2 as positive controls. We use TMAs of tonsillar tissue from many different tonsils for optimizing hematopoietic markers and for validation. Other TMAs serve as negative controls and for validation. We do not include positive controls on the same slide as patient tissue; instead, we use one separate control for several cases (unless they come from a separate facility, in which case a positive control is sent with patient slides). We have attempted to cut positive controls at one end of the slide, but so far have not built up an inventory (28).

I remember, for many clinical research studies, having to buy antigen and then assess how much to use to adsorb the antibody, so that we had an adequate negative control to satisfy journal reviewers. This was assumed to be better than the serum or buffer we would otherwise have used as a negative control. Today, in the age of cost containment and standardization, we do not use negative controls with patient specimens, although they are used for optimization and research projects. This makes for a shorter run or more patient slides stained per run and reduces cost. When slides are stained in a panel, each can act as a negative control for the others. However, I clearly remember staining kidney frozen sections with the peroxidase/antiperoxidase technique and being delighted with the excellent results – only to find that our negative controls stained just as beautifully.

Antibody cocktails

We also perform IHC using cocktails (29,30,31,32), which is of immeasurable use in defining difficult-to-diagnose cancers. An example is PIN4, a cocktail of antibodies against high-molecular-weight keratin, p63, and p504S that we use to diagnose prostate cancer in needle biopsies. Cytoplasmic keratin and nuclear p63 are negative because basal cells are absent, whereas p504S (AMACR, racemase) stains the luminal cells. Although we have used commercial cocktails for this purpose, we now make our own with excellent results. We also use homemade cocktails of TTF1 and napsin A for lung adenocarcinoma; cytoplasmic high-molecular-weight keratin and nuclear p40 or p63 for squamous cell carcinoma; tumor pan-cytokeratin and CD31 for angiolymphatic invasion; and HMB45, melan A, and tyrosinase for malignant melanoma. By combining any two of our working antibodies, we can create a good cocktail for a particular use (30,33).

Cytology cell blocks

We perform IHC on cell blocks of fine needle aspirations (FNAs) and fluids, as well as surgical specimens, such as resections and biopsies (34). Results are similar and excellent – even after the use of a rapid processor for biopsies. We’ve only encountered problems with certain alcohol-based fixatives and transport media, but using formalin has improved the situation.

In situ hybridization

We can now do ISH on our automated stainers for kappa and lambda, human papillomaviruses 6, 8, 16, and 18, cytomegalovirus, and Epstein-Barr virus (35). For kappa and lambda, ISH is much more sensitive than IHC because of the absence of background; for human papillomaviruses it is also much more sensitive – but only as individual viruses, not cocktails; cytomegalovirus, in contrast, is better by IHC. Other organisms that stain well by IHC include adenoviruses, Toxoplasma, Helicobacter, Treponema pallidum (no more time-consuming searches), herpesviruses 1 and 2, and polyomavirus. Other viruses, such as high-risk papillomavirus, also stain well with RNA ISH using RNAscope and probe.

Carcinoma of unknown origin

We can now use IHC to identify site of origin of carcinomas of unknown primary (36,37), which allows us to help clinicians determine treatment approaches – for instance, if biomarkers of a particular cancer arise, if predictive markers are present, or as a companion diagnostic to suggest that a patient may respond to a particular treatment (for instance, anti-PD-L1 therapy) (38).

Molecular markers

IHC can be used as a surrogate for molecular testing. Molecular markers we have studied by IHC include MSI (39), wherein five Barrett’s esophagus-associated adenocarcinomas showed loss of MLH1 and PMS2 expression; all of those showed high-level MSI by PCR and four showed hMLH1 promoter methylation. IHCwas 92 percent sensitive and 89 percent specific compared with MYC rearrangement by FISH in 31 lymphomas (40), and ALK was 100 percent sensitive and 96 percent specific in 110 non-small cell lung carcinomas (41). Molecular IHC surrogates are also useful in breast cancer (42,43).

Post-analytic parameters

Image cytometry

IHC quantitation can be performed by visual assessment of intensity, percent positive, and distribution or by image cytometry, which eliminates the human element. Over the years, we have used the CAS 200 (Dako), ACIS Chromovision (Dako), and now Aperio (Leica) to quantitate breast markers, but we still review each slide visually as well to ensure that the correct areas are quantitated and results are similar by both methods. Other situations where image cytometric quantitation with digitization are used include MIB1 proliferation in breast carcinoma, brain, and neuroendocrine tumors – although some feel the latter is best quantitated manually (44). Thus, the qualitative method has become quantitative, requiring even more standardization and validation.

Whole-slide scanning and digital imaging

We already use digital imaging for a plethora of reasons: photography of gross specimens, reports, microscopic pictures used in teaching, tumor registry meetings, sharing of cases, and on a daily basis for quantitating breast cancer IHC markers and ordering IHC with generation of a daily list. Sections are still cut, stained/immunostained, and delivered to residents and pathologists, but the workflow is now documented by scanning barcoded slides on their journey from gross room to laboratory to microscope. We do our immunostaining on automated machines, each method documented and controlled by computer (46). Thus, we are moving closer to the Lean laboratory of efficiency, quality, and high productivity. We performed image analysis for ER quantitation 25 years ago (45), but we never dreamed we would be running like an automated Toyota Corporation (47) today – the lab of the future!

In the early days, we had to ask a photographer to take pictures of microscopic slides and wait at least two days for the resulting kodachromes (48). Later, we purchased a Lasergraphics film scanner and could deliver the whole film, taken with a microscope camera and processed by the Lasergraphics machine, to the camera store. That took only a single day to develop. Now, the digitized microscopic image is photographed and available immediately online for a poster, talk, registry meeting, or as a print for a manuscript. It used to take three months for the residents to assemble their pathobiology talks; now, they are done in days. Speaking of talks, we used to have to ask the photography department to make diazo kodachromes (49) of our typed descriptions, all of which went into a carousel and often stuck during presentations. No such problems now… as long as the computer functions well!

Whole-slide imaging received FDA approval in April 2017 for primary diagnosis (50) and it has taken off ever since. I recently saw a webinar where the imaged IHC slide was viewed on the monitor together with the H&E images for easier diagnosis!

Companion diagnostics

Increasingly, companion diagnostics are becoming a valuable application of IHC (51). Pembrolizumab was initially FDA-approved for use with metastatic – and now also primary – non-small cell lung carcinoma (NSCLC); PD-L1 as a companion antibody shows strong intensity and >50 percent of cells are immunopositive (52). Pembrolizumab is also FDA-approved as a complementary diagnostic in metastatic melanoma and PD-L1 and nivalumab in NSCLC, metastatic melanoma, Hodgkin’s lymphoma, and urothelial, renal cell, and head and neck carcinoma independent of tissue PD-L1 expression (52,53). Two FDA-approved PD-L1 antibodies are now available from Dako (clones 22C3 and 28-8), but at high cost of $3,500–4,000 per 50 tests. Other antibodies from different companies, with different cutoff values and for use with different carcinomas and therapeutic inhibitors, are also available (54,55).

Techniques

Techniques that may not yet be fully integrated into clinical anatomic pathology include miRNA, RNAscope for RNA ISH (with very expensive probes, but a large selection), and molecular methods (56), many of which can be done quicker and cheaper by IHC, MSI, and digital imaging with whole-slide scanning.

We have done technical validation studies using RNA ISH for 18 high-risk papillomaviruses. 100 percent of p16 IHC-positive and 50 percent of p16 negative head and neck squamous cell carcinomas were RNA ISH-positive with 88.9 percent concordance with p16 IHC (13). We obtained excellent, easy-to-read results (see Figure 3). Our EBER and CMV ISH studies showed 90.3 and 66.7 percent concordance with their IHC equivalents, respectively. Eight of 16 (50 percent) negative CMV IHC were positive by RNA ISH (57).

It seems that, for viruses, RNA ISH is a valid alternative to IHC – but concordance for cancers is much lower, particularly in the case of PD-L1 testing. Whereas 90 percent of lung adenocarcinomas and no metastatic colon carcinomas were PD-L1-positive by IHC, 60 percent of the same lung cancers and 25 percent of colon cancers were PD-L1-positive by RNA ISH (58). For other applications, RNA testing proved more reliable; TTF1 and Napsin A IHC and RNA ISH were both 95 percent sensitive and 100 percent specific (37). Albumin RNAscope was 100 percent sensitive and specific for hepatocellular carcinoma (59), but TFE3 RNAscope was not a viable alternative to IHC or FISH (60).

The capture of circulating tumor cells and their genetic material has shown progress with improved technical and sequencing-based methods, bringing the possibility of liquid biopsy of solid tumors closer to reality. However, its clinical utility for diagnosis and treatment is still unproven (61).

Detection of protein in formalin-fixed, paraffin-embedded (FFPE) tissue by IHC is semi-quantitative. The possible future use of protein biomarkers detected by proteomics combined with IHC from FFPE tissues is exciting. A novel amplification system enables quantification of protein by counting dots (62) and can be combined with IHC. Combining IHC with mass spectrometry in the same tissue section allows highly multiplexed IHC (using three or more monoclonal antibodies) for direct quantitative imaging (63).For HER2, results from this approach were comparable with or better than the reference methods of ELISA and flow cytometry.

I can only imagine, with further technological advances, what the next decade – let alone four – will bring to the diagnostic laboratory. We are already in line for diagnostic whole-slide imaging, with FDA approval in the United States and several laboratories in Europe already using it. This digital approach will include scanning of all IHC-stained slides – unless, that is, we go entirely to molecular testing. Personally, I don’t see that as a possibility; I believe that anatomic pathologists will still want to see the tissue in H&E and immunostained slides, whether on a computer monitor or through a microscope.