A Novel Quality Control Slide for Quantitative Immunohistochemistry Testing

Abstract

We introduce a novel quality control technology that may improve intra- and interlaboratory immunohistochemistry (IHC) standardization. The technology involves the creation of standardized antibody targets that are attached to the same slides as the patient sample. After IHC staining, the targets turn the same color as the stained cells or tissue elements. Unlike current clinical practice, our proposed targets are neither cells nor tissue sections. To create reproducible standards that are available in unlimited supply, we use short constrained peptides as antibody targets. These peptides are attached directly to the glass slide. We show that these peptides simulate the portion of the native antigen to which the antibody binds. They are useful in detecting subtle changes in IHC staining efficacy. Moreover, the peptides do not degrade after deparaffinization or antigen retrieval treatments. This technology may be valuable in creating nationally standardized controls to quantify IHC analytical variability.

When immunohistochemistry (IHC) was first introduced as an adjunct in surgical pathology diagnosis, the interpretation of tissue staining was largely qualitative. Specific markers were present or absent, thereby characterizing a tumor cell's lineage. The fact that IHC interpretation was qualitative, rather than semi-quantitative, minimized the impact of many known inconsistencies among laboratories with regard to reagents and methods. Early on, it was recognized that quality control, reproducibility among laboratories, and standardization were important issues (Elias et al. 1989; Rickert and Maliniak 1989; Wold et al. 1989; Bosman et al. 1992; Taylor 1992). Published studies based on early proficiency test surveys found that the overwhelming majority of clinical IHC laboratories were able to obtain accurate qualitative results, i.e., to identify samples as either positive or negative (Wold et al. 1989; Bosman et al. 1992).

During the past several years, there has been a rapidly escalating clinical need to perform IHC stains that require quantitative interpretation. The level of cellular expression for certain analytes, notably HER-2 and estrogen and progesterone receptor proteins, are linked to particular therapies. In this emerging clinical paradigm of individualized medicine (Elledge and Osborne 1997), accurate quantitative data are important to reach the correct treatment decision. A look into the pharmaceutical pipeline suggests that other future anti-cancer drugs will be similarly dependent on accurate quantitative IHC staining and interpretation. The growth of quantitative IHC, however, has raised the bar for accuracy and reproducibility.

Recent studies indicate that interlaboratory reproducibility for quantitative IHC tests, such as HER-2 and ER, requires improvement (Nicholson and Leake 2000; Rhodes et al. 2000a,b). Reproducibility in quantitative IHC testing is far below that normally found among other clinical laboratories, such as those in clinical chemistry (Leake et al. 2000; Nicholson and Leake 2000; Rhodes et al. 2000a, b). Previous efforts towards IHC standardization and reproducibility have generally sought to promulgate standardized assay methodologies (Forrest and Barnett 1989; Rickert and Maliniak 1989; Balaton et al. 1996; National Committee for Clinical Laboratory Standards 1997; Williams et al. 1997; Leake et al. 2000), develop objective methods of measurement (Cross 2000; Esteban et al. 1993), and provide external staining standards (Wold et al. 1989; Lambkin et al. 1998; Bosman et al. 1992; Rhodes et al. 2000a, b). External standards (e.g., tissue sections from a central laboratory) are part of national surveys, such as those managed by the CAP in the United States and the NEQAS program in the United Kingdom. External proficiency tests are important aspects of a total approach towards quality assurance. However, the absence of validated controls for daily assay verification is a serious handicap in fostering standardization and reproducibility. This is in contrast to other types of clinical laboratories, such as clinical chemistry or hematology laboratories, which have validated controls to check reagents, methods, and instrumentation. Laboratory staff plot the data generated from the controls in the form of a Levey-Jennings chart (Westgard et al. 1981).

By contrast, clinical IHC laboratories are instructed to establish their own controls for daily use (National Committee for Clinical Laboratory Standards 1997). Because each control is homegrown and is verified by the same in-house test for which it is supposed to act as a control, the controls are inherently unstandardized. Levey-Jennings charts for tracking daily performance of quantitative stains are virtually unknown in the clinical IHC laboratory. Whereas manufacturers generate large lots of serum calibrators and controls capable of supplying the daily needs of thousands of clinical chemistry laboratories, the same cannot be accomplished with tissue biopsies. The biological resource is rapidly exhausted and not reproducible. Moreover, it is not standardized.

In summary, quality assurance methods that are considered routine practice in other clinical laboratories are not found in the clinical IHC lab. This has been unavoidable because the nature of the test presents unique challenges in implementing comparable quality assurance methods. Therefore, there is a need for better quality control technologies that can provide constant daily feedback to clinical IHC laboratory staff. Ideally, these technologies will be quantitative and will provide an early indication as to when an IHC test is drifting out of range.

Here we present a new technology that can, in part, address these quality assurance needs. Our technology provides an inexhaustible, reproducible, quantitative, antigen-specific, stable, and inexpensive source of analytical control material. By design, the controls are not affected by any pretreatment steps, such as antigen retrieval. Therefore, it measures the efficacy of analytic components of the IHC stain to a level of precision and standardization not previously possible. Moreover, it provides a simple and rapid indicator to show if an outright mistake occurred in the IHC stain, such as if the wrong reagent had been placed on a tissue section.

Materials and Methods

Antibodies

Controls for the following monoclonal antibodies (MAbs) are described in this study: anti-PR (clone 636), anti-ER (clone 1D5), anti-p53 (clone DO7), and anti-Ki-67 (clone MIB1), all of which were supplied through CytoLogix Corporation (Cambridge, MA).

Immunohistochemical Staining

All IHC stains were performed on formalin-fixed and paraffin-embedded antigen-positive human tissues. Serial tissue sections were deparaffinized in xylene, dipped in decreasing concentrations of ethyl alcohol, and then rehydrated in water. Epitope retrieval was then performed by incubating the slides in a pressure cooker (Nordic Ware; Minneapolis, MN) for 30 min in a 0.01 mol/liter concentration of citrate buffer (pH 6.0). Slides were immunostained either manually or with an automated slide stainer (Artisan, CytoLogix Corp.) using the company's IHC detection kit, a labeled streptavidin–biotin detection system.

ER antigen-specific primary MAb 1D5 (Al Saati et al. 1993) was used at the standard concentration supplied by CytoLogix (∼3 (μg/ml). The antibody was added to the tissue section and incubated at room temperature (RT) for 40 min. PR clone 636, Ki-67 clone MIB-1, and p53 clone DO7 were used at the working concentration supplied by the manufacturer (CytoLogix Corp.). In contrast to the conditions for ER staining, those MAbs were applied to tissue sections at 37C for 20 min. The slides were then rinsed and the presence of bound primary antibody (to the tissue antigen) was detected using the CytoLogix IHC detection kit. Briefly, these steps include incubation with biotin-conjugated horse anti-mouse IgG (heavy and light chain-specific) secondary antibody for 20 min at RT. Next the slides were incubated with horseradish peroxidase-conjugated streptavidin for 20 min at RT. The color was then produced with liquid-stable DAB (3, 3′-diaminobenzidine tetrahydrochloride)/hydrogen peroxide for 10 min and enhanced with 5% (w/v) copper enhancer (cupric sulfate pentahydrate) for 10 min. For slides that were used for image quantification, no counterstain was applied so as to simplify image colorimetric quantitation.

Application of Peptides to Glass Slides

Peptides (as described in Sompuram et al. 2002) were covalently coupled to the isocyanate-derivatized glass surface of microscope slides. Briefly, 1 μl of various peptide concentrations was spotted onto activated, isocyanate-derivatized slides. The peptides were allowed to covalently couple to the glass surface (via the ∊-amino group of a C-terminal lysine) for 15 min. The slides were then rinsed and remaining reactive isocyanate groups were quenched with bovine γ-globulins (0.05%; Sigma, St Louis, MO).

Peptide Specificity and Crossreactivity

PR, Ki-67, and p53 peptides were tested for binding to different mouse MAbs or to a polyclonal antibody. Eighteen spots of each of the peptides were applied to an isocyanate-activated microscope slide. Each spot comprised approximately 20 picomoles in a 1-μl volume. Various antibodies were then reacted with the spots: estrogen receptor (ER) clone 1D5, progesterone receptor (PR) clone 636, p53 clone DO7, Ki-67 clone MIB-1, vimentin clone V9, leukocyte common antigen (CD45), cytokeratin clones AE1 and AE3 (cocktail), melanocyte-specific antibody clone HMB 45, S100 polyclonal antibodies (all per the manufacturer's supplied concentration; CytoLogix), progesterone receptor clone 1A6 (DAKO; Carpinteria, CA), myeloma protein MOPC 141 and mouse polyclonal IgG (both at 1 μg/ml from Sigma). Each spot was separated from the others on the slide by drawing hydrophobic barriers with a Pap pen (Research Products International; Mount Prospect, IL). This prevented cross-contamination of the reagents from one spot to the others. After incubation with the primary antibodies, all 18 spots were then developed with a Universal IHC detection kit/DAB (CytoLogix).

Colorimetric Intensity Measurement of Peptide Spots

The colorimetric intensity of the IHC-stained peptide spots was obtained by scanning the slides with a flatbed scanner (Perfection 1200U; Epson America, Torrance, CA). The image was stored in Adobe Photoshop. The color intensity of the spots was then quantified using a scion image program (Scion Corporation; Frederick, MD). Peptide spot color intensity is expressed as mean pixel optical density on a 1–256 scale.

Inhibition of MAb Binding by Peptide

We tested the ability of the peptides to inhibit the binding of antigen-specific MAbs to native antigen. As a source of native antigen, we used tissue sections containing cells that strongly express the antigen in question. If a peptide inhibits the binding of an MAb to native antigen, then that is detectable as inhibition of IHC staining on the tissue section. Aliquots containing the antigen-specific MAb were incubated at 37C for 45 min with various concentrations of peptides. The MAb was then used as a primary antibody in an IHC staining assay. The IHC staining assay was performed as described above, in triplicate on serial tissue sections. The intensity of tissue staining was quantified by a microscope-based image analysis program (Image Pro Plus; Media Cybernetics, Silver Spring, MD).

Reagent Failure Simulation

To test the utility of the controls, we investigated the decrement in color intensity on the control spots and tissue sections under reagent failure conditions. We were primarily interested in determining how much degradation a particular reagent would need to experience before a consistent decrement in color intensity would be detected. We simulated failure by diluting the primary antibody. In separate experiments, we also diluted the secondary anti-mouse IgG–biotin conjugate or the streptavidin–peroxidase conjugate. Briefly, tissue sections were mounted on slides that previously had been spotted with peptides. Primary antibody (or secondary antibody or tertiary reagent) was serially (1:2) diluted and used for IHC staining as described above. Tissue staining intensity was quantified by a microscope-based image analysis program (Image Pro Plus). Peptide spot color intensity was measured as described above.

Sensitivity of Peptide-coated Slides to Baking, Xylene, Alcohol, and Boiling

Slides are routinely subjected to these treatments in preparing paraffin-embedded, formalin-fixed sections for IHC analysis. Therefore, we tested the sensitivity of peptide-spotted slides to these treatments. Baking tests the sensitivity of the peptide-coupled glass slides to dry heat. Slides are baked in a 60C oven for 30 min to 1 hr. The xylene and alcohol treatments are part of the standard deparaffinization process. Slides are successively dipped in three Coplin jars of xylene for 3 min each. The slides are then immersed in a series of three Coplin jars of absolute ethanol, followed by two Coplin jars of 95% ethanol and two Coplin jars of 70% ethanol, all for 2 min each. Tissues are then rehydrated in water. Epitope retrieval is then performed by incubating the slides in a pressure cooker (Nordic Ware) for 30 min in a 0.01-mol/liter concentration of citrate buffer (pH 6.0).

Results

Technology Concept

An important goal of this project was the identification of antibody targets that would be available in unlimited supply, easy to manufacture or purify, inexpensive, antigen-specific, and stable after tissue pretreatment steps (deparaffinization and antigen retrieval). Small peptides can have these properties. Moreover, small peptides can mimic the portion of the native protein to which an MAb binds. The concept is illustrated in Figure 1. An antibody can bind equally well to the native protein as to the peptide that mimics the native antigen (“synthetic peptide”). We identified a series of peptides that mimic the antibody-binding region of the human estrogen receptor (ER) clone 1D5, progesterone receptor (PR) clone 636, Ki-67 clone MIB-1, p53 clone DO7, and HER-2 clones 9C2 and 11G5. We describe peptide identification by phage display and affinity measurements elsewhere (Sompuram et al. 2002).

Figure 1

Schematic showing the relationship of the native protein to the peptide mimic. The peptide represents the antibody-binding epitope of a complex protein. The epitope can represent a linear sequence of the native protein. Alternatively, the epitope can be formed by amino acids that are not immediately adjacent to each other in the primary sequence. Rather, the epitope might be formed by non-contiguous amino acids that are brought together by the three-dimensional folding of the protein.

Figure 2

Demonstration of the linear range for the PR MAb IHC stain against various concentrations of PR peptide. One microliter of various concentrations of PR peptide was coupled to glass slides and subjected to the standard IHC staining procedure. Peptide spot intensity is measured as mean pixel intensity on a 1–256 scale.

Linear Assay Response

As a first step in validating the utility of specific peptides as quality control targets, we first sought to identify the linear range of concentration for each peptide mimic (i.e., PR, Ki-67, and p53 peptides). To use peptides as IHC controls, we first attached the peptides to glass slides (per the protocol described in Materials and Methods). We covalently coupled various concentrations (doubling dilutions) of each synthetic peptide to glass slides, to identify a linear range for the stained peptide spots. We found that the peptides yielded linear dose–response curves in approximately the 1–20 picomole/μl range. Representative data for the PR peptide are shown in Figure 2.

Peptide Targets Mimic Native Antigen

We then tested the specificity of each peptide, coupled to a glass slide, for the relevant MAb. If the peptides accurately mimic an epitope of the native antigen, then they will only react with the relevant antibody. No other MAbs should react with the peptide target. We examined this hypothesis by testing the reactivity of a panel of MAbs for each selected peptide. Using a wax pencil, we first drew a 6 × 3 matrix of lines on glass slides, creating 18 fluid distinct zones (Figure 3). We then spotted an identical peptide into the center of each zone. By doing this, each peptide is separated from the next by a hydrophobic barrier. Each spot was then reacted with a primary antibody, comprising one of 15 MAbs, a polyclonal antibody, or the relevant peptide-specific MAb. Antibody crossover was prevented by the hydrophobic barriers. Binding of the primary antibody to a peptide spot was detected by standard IHC detection methods (see Materials and Methods). As shown in Figure 3, only the PR 636 MAb bound to the PR peptide spot while the Ki-67 MAb bound the corresponding Ki-67 peptide spot. In all cases the two relevant spots, positioned at opposite corners of the slide, were stained with no detectable crossreactivity to other MAbs (arrows, Figure 3). The same was true for other peptides as well, such as the ER peptide, described elsewhere (Sompuram et al. 2002) and the p53 peptide (data not shown). These data show that each peptide is bound only by the relevant MAb.

We further examined the fine specificity of the peptide for its cognate MAb by determining if the peptide binds at the antibody's antigen-binding site. If a peptide binds at the antigen-binding site, then it should inhibit the binding of the antibody to its native antigen. We tested this hypothesis with an inhibition assay. For native antigen we used tissue sections containing the target antigen. For example, PR specificity analysis was accomplished using breast carcinomas that express high levels of PR receptor. Various concentrations of each peptide were pre-mixed with an MAb. The MAb was then applied to tissue sections as a primary antibody (see Materials and Methods). Primary antibody without any peptide inhibitor represented the positive control, while other irrelevant peptides were used as negative controls. As shown in Figure 4 (A), the PR-specific peptide inhibited the binding of the PR MAb to PR receptor in a dose-dependent fashion. Other peptides, at even higher concentrations, did not inhibit the interaction of the antibody with native antigen. Figure 4 also shows specificity data for the Ki-67-specific peptide (B). It, too, was able to inhibit the binding of the Ki-67 (MIB 1 clone) MAb to its target antigen (in sections of tonsil). These data indicate that the peptides mimic the epitope of the native antigen and bind to the antigen-combining site of the MAb. Similar data were obtained for other peptides as well, such as the ER peptide, described elsewhere (Sompuram et al. 2002), as well as the p53, and HER-2 peptides (data not shown).

Figure 3

The peptides, coupled to glass slides, mimic the antigenic epitope. One-microliter spots of PR (left panel) or Ki-67 (right panel) peptides were placed in the center of each grid location. Various antibodies and controls were then applied to the different grid locations. The specific antibodies for each peptide mimic are applied at opposite corners of the slide (arrows). The legend to the left of each slide describes the antibodies or controls that were applied to each grid location (see Materials and Methods).

Figure 4

Peptide-specific inhibition of antibody–antigen interactions. Stained tissue images were quantified and plotted as optical image density (y-axis) against the inhibiting peptide (x-axis). The units for image density on tissue sections are optical density on a 0–2 scale. A shows the staining intensity of PR antibody on PR⁺ tumor cells after pretreatment with various peptides. As shown in A, only the PR-specific peptides inhibit the antibody–antigen interaction. B follows a similar format but relates to staining of Ki-67⁺ cells with a Ki-67 MAb.

Figure 5

Peptide spot color intensity as a function of doubling dilutions of primary (PR) antibody. PR peptides and PR⁺ tissue sections were both placed on the same slides and stained with various dilutions of the PR MAb. Color intensity of the peptide spots (square symbols) or tumor cells (triangle symbols) was measured and plotted on the y-axis. The figure shows a linear decline in intensity with decreasing antibody concentrations for both the peptide spots and the tissue sections. Tissue color intensity is measured as optical density on a 0–2 scale. Peptide spot color is measured as mean pixel intensity on a 1–256 scale.

QC Indicators’ Sensitivity to IHC Stain Degradation

Having verified the specificity of the peptides, we then tested their utility as quality control (QC) indicators. Ideally, a QC indicator will provide an early warning of newly developing problems so that a correction can be implemented before it affects patient care. In a first set of experiments, we asked if the peptides can be used to create a sensitive indicator of reagent deterioration or abnormally functioning instrumentation. Both of these circumstances can result in an abnormally low amount of antibody being applied to the tissue section. For example, if an antibody deteriorates, then an abnormally low amount of functioning antibody is applied to the tissue. Instrument malfunctions can also affect stain performance, such as through insufficient reagent dispense volume, inappropriate temperature, or evaporation of reagent. Applying suboptimal concentrations of antibody in an IHC stain can, at least in part, simulate all of these conditions. Therefore, we reasoned that diluting reagents in the IHC stain could simulate these failure modes.

We tested the QC indicators’ ability to detect subtle levels of reagent failure by simulating failure in each of the three components of the IHC assay, i.e., primary antibody, secondary antibody, and streptavidin–peroxidase conjugate. The results are similar regardless of which reagent is diluted (data not shown). Representative data for primary antibody (PR-specific) dilution on breast carcinoma are shown in Figure 5. Each data point represents a triplicate measurement of IHC stain intensity, as measured on the tissue section (by image analysis) or on the peptide spot (by scanning). We find an essentially linear decrement in stain intensity from dilutions of 1:4–1:64. As shown in Figure 5, the decrement is equivalently reflected in both tissue and peptide spot optical density. These findings are representative of those found with other peptides and antibodies as well.

It is our experience that fine decrements are difficult to detect by eye. In our estimation, decrements in tissue staining were not obvious until the dilutions approached the 1:16 range (Figure 6). The precise sensitivity to partial reagent or instrument failure probably depends on the starting concentration of the antibody and the amount of analyte in the tissue section. For these studies we used the manufacturer's recommended concentration on a strongly PR⁺ tissue. Figure 6 shows representative photomicrographic images of stained PR⁺ tumor cells used for measuring image optical density. Without the aid of image analysis, we found it nearly impossible to distinguish two- or fourfold dilutions of antibody. Without a side-by-side comparison to the positive control (undiluted) group, even larger antibody dilutions could be hard to identify when judged by eye, especially when the decline occurred gradually. These issues could, in part, be addressed using image analysis, as we did in Figure 5. However, we would still lack a source of standardized tissues that contain a defined level of antigen. The use of peptides having a defined molar amount of analyte on the glass addresses this shortcoming. A potential advantage of the peptide spots is that they do not require image analysis for quantification. We are developing a simple and inexpensive slide densitometer that would facilitate easy quantification of stained spot color intensity.

QC Indicators Are Stable

If the peptides coupled to glass slides are to ultimately be a practical quality control product, then they must be stable to the routine pretreatments that tissue sections undergo. For example, slides with paraffin-embedded tissue sections are baked at 60C for about an hour to promote tissue adherence. Tissues are also immersed in xylene and ethanol for deparaffinization. Lastly, most tissue sections are boiled in pressure cookers as part of antigen retrieval. It is important that any quality control indicator be capable of withstanding these treatments without denaturation. One or more of these treatments would normally be expected to denature many larger, more complex proteins. We tested each of these treatments and found that small, constrained peptides are extraordinarily robust. Figure 7 demonstrates representative data for IHC staining of peptide spots with and without baking (60C dry heat, 1 hr) of the peptide-coated slides. We consistently failed to see any noticeable decrement in the staining signal after baking. Figure 8 illustrates our findings with respect to deparaffinization and antigen retrieval. For this particular study, we compared an ER peptide (crosshatched bars) to native murine IgG protein (white bars). The ER peptide is detected by the ER 1D5 MAb. The mouse IgG protein is detected by the anti-mouse IgG–biotin conjugate, the secondary antibody in the staining process. The data show that the ER peptide is not affected by the deparaffinization and antigen retrieval processes. The peptide spot intensity is essentially unchanged after the treatment. This was true for all of the peptides tested (data not shown). Figure 8 also illustrates that, in contrast to short constrained peptides, large globular proteins such as IgG are somewhat susceptible to these treatments. After xylene/ethanol and boiling (antigen retrieval) treatments, there is a consistent but mild decrease in staining intensity of IgG spots. Presumably, some of the epitopes recognized by the polyclonal anti-mouse IgG–biotin conjugate are lost after boiling due to protein denaturation.

Figure 6

PR⁺ tumor tissue photomicrographs after staining with doubling dilutions of PR antibody. The antibody dilution (beyond the standard working concentration) is indicated at the top of each panel. This image demonstrates that it can be difficult to ascertain, by eye, subtle degradation of a reagent solely on the basis of microscopic images.

Figure 7

Peptides do not denature after baking (dry heat). Peptide-coupled slides were treated (as indicated on the x-axis) and then immunohistochemically stained. In this particular example, an ER peptide with an ER MAb was used. The resulting peptide spot intensity (mean pixel intensity on a 1–256 scale) was measured and is shown on the y-axis.

Figure 8

Peptides do not denature after treatment for deparaffinization and antigen retrieval. ER peptides (crosshatched bars) or murine IgG (white bars) on glass slides were given the indicated treatment (x-axis categories). Peptide spot intensity (mean pixel intensity on a 1–256 scale) after IHC staining is shown on the y-axis.

Discussion

Standardization of quantitative IHC has been hampered by a lack of internationally agreed-on reference standards and calibrators. The current practice of using pathological discard tissue is suboptimal because each patient's biopsy material differs from other patients’ biopsies. This lack of reproducibility makes it difficult to reliably compare the staining processes from different IHC laboratories. In an effort to rigorously address this need, we sought to develop a technology that would be integrated into the microscope slide so that there would be an on-slide control with every IHC test. In embarking on this project, we first established a set of criteria against which any technology might be judged. Those criteria are that the ultimate product should be (a) a highly reproducible and quantitative standard, (b) available in unlimited quantities, (c) antigen-specific, (d) inexpensive, and (e) stable.

A few reports have proposed that cell lines, mounted on slides, be used as controls (Allred 1993; Riera et al. 1999). Cell lines can be grown in vitro, formalin-fixed, and embedded in paraffin or agarose for sectioning and mounting on glass microscope slides. Cell lines offer the advantage of being potentially available in unlimited quantities, addressing the second criterion. This feature overcomes a critical limitation of tissues as controls. For an individual laboratory, cell lines may work well as an intralaboratory standard. However, as a national QC standard, using cell lines in such a manner suffers from problems relating to manufacturability, cost, and consistency. In general, the amount of an analyte in a transformed cell line is somewhat heterogeneous, depending on cell cycle and growth conditions. Consequently, different batches of cells may not express consistent levels of the analyte. In addition, the amount of any particular analyte expressed by the cells can drift over time. This phenomenon has been attributed to the outgrowth of subclones expressing higher or lower amounts of the analyte. Therefore, a manufacturer will need to periodically subclone the cells to force them to express a predefined amount of analyte. This need further complicates their manufacture and increases the cost. If a manufacturer were to section the blocks and provide them premounted on slides, costs would rapidly escalate because of the labor-intensive nature of histological sectioning. Leaving the sectioning and mounting to customers simply shifts the labor cost burden to the laboratories, many of which are already understaffed.

These considerations led us to conclude that the most practical type of national QC standard would be the purified analyte itself. The only problem to overcome would be that most of the analytes are present in only trace amounts from cells or tissues. Producing them through recombinant DNA methods is certainly possible, but not in a way that would satisfy our fourth criterion, relating to cost. Each analyte would likely present its own production and purification problems. These production considerations are acceptable in the cost context of pharmaceuticals but not as IHC quality control material. To overcome this last hurdle, we landed on the concept of using only that portion of the native analyte to which the antibody binds as the QC target. The target is a synthetic peptide that resembles the three-dimensional conformation of the epitope to which the antibody binds. Short peptides can be synthesized in large quantities and quite cheaply. In fact, preliminary cost analysis of this technology, if it were to be commercialized, showed that the manufacturing costs associated with the peptide are negligible. Rather, the costs would be driven by the process of chemically activating the glass and applying the peptide spots. This process could ultimately be handled robotically if there were sufficient demand to justify the investment.

The technology we developed provides for the application of a precise molar amount of analyte to the glass slide, thereby creating an exact, reproducible quantitative standard. Because the tissue sample is mounted on the same slide as the control indicators, both are treated in an identical fashion. As the tissue sample is stained, so too are the control indicators. The precise amount of color that develops on the synthetic controls directly reflects the efficacy of the IHC stain's analytic components. We have also developed a novel slide densitometer to quantitatively and easily measure the optical density of the controls. In a modified format, the same technology can be coupled with existing quantitative tissue imaging systems.

High-affinity antigen mimics (peptides) are identified using phage display technology. Phage display is a combinatorial library technique that enables us to rapidly screen approximately 10⁸–10⁹ different peptide combinations. Phage display allows us to identify antigen mimics without any prior information about the antigen itself. We describe the derivation of peptides in a separate report (Sompuram et al. 2002).

These synthetic peptides are covalently attached to the glass slide using an isocyanate coupling chemistry. In general, small peptides do not attach well to solid-phase matrices. To efficiently and rapidly attach peptides to glass slides, we developed a coupling chemistry based on isocyanate-reactive groups. Isocyanate groups are highly reactive and attach to the peptide through an ∊-amino group on a terminal lysine of each peptide. The chemical attachment method has the secondary benefit of being highly reactive not only to the peptides but also to tissue proteins in general. This coating may therefore facilitate firm attachment of tissue sections to the glass. For example, isocyanate groups are vastly better binders of peptides than are slides coated with poly-l-lysine, a common coating for improving tissue section adherence. In the future, we plan to examine whether this chemical reactivity translates into improved adherence to tissue sections.

We have demonstrated here that the peptides behave as would be expected of native proteins. They produce a normal linear dose–response curve (Figure 2), are antigen-specific mimics of native proteins (Figures 3 and 4), and simulate analytical errors in a fashion that would be expected of the native proteins themselves (Figure 5). Without precise quantitation, such analytic errors would be difficult to detect early on by eye, especially if the decline was gradual (Figure 6). The peptides and coupling chemistry are so robust that they are not affected by any of the pretreatment steps associated with deparaffinization and antigen retrieval (Figures 7 and 8).

This technology has two drawbacks that we can perceive. The most important is that it is sensitive only to analytic errors, such as those associated with reagents, instrumentation, and technique. The peptides (at least in their current configuration) are not susceptible to pre-analytic treatments, most importantly fixation and antigen retrieval. This issue is particularly important because pre-analytic sources of error are greater for IHC testing than for other conventional laboratory tests, such as those that measure blood analytes. There is probably no routine pretreatment of blood that quite matches the level of error that can be introduced by fixation and antigen retrieval. Therefore, tissue sections will reflect variables associated with fixation and antigen retrieval (pre-analytic aspects) as well as reagents, instrumentation, and methods (analytic components), and will thus reflect a broader range of staining variables. For this reason, we do not propose (at this time) the complete replacement of tissue sections as controls. Tissue sections (as controls) represent a check on antigen retrieval. It is our view that tissue sections do not control for fixation errors, because the fixation of the control tissue generally has no relationship to possible fixation errors associated with a patient sample. If errors in fixation were to occur on a particular day, they would not be reflected in a control tissue sample that was fixed and embedded days, weeks, or months before the day in question.

The other disadvantage of this technology is that it is specific to a particular MAb. This disadvantage is inherent in the strength of the technology. Using peptides as analytical targets causes these QC indicators to be inexpensive, available in unlimited supply, antigen-specific, and stable in the face of deparaffinization and antigen retrieval. Fortunately, the task of creating controls for each antibody is not as high a hurdle as it may appear. The need to use antibodies that recognize their targets after formalin-fixation has narrowed the field of clinically useful MAbs. Consequently, for each analyte there are relatively few reliable MAbs that are commonly used for clinical purposes. With that in mind, we made controls for popular, widely used MAb clones.

Quantitative IHC controls will be important for quantitative IHC tests. At present there are only a handful of such tests. ER, PR, and HER-2 are probably the most important. As more quantitative IHC tests are introduced, the need for better quality assurance methods will continue to grow. Over time, we anticipate that the IHC laboratory will adopt techniques that are a standard of practice in other clinical laboratories. For example, we believe it only a matter of time until clinical IHC laboratories begin to use Levey-Jennings charts, commonly used in other clinical laboratories. Here we describe a new analytical tool that can be incorporated into a total quality approach. It is unique in that it is quantitative and distinguishes analytic from pre-analytic sources of error. To our knowledge, there are no published survey studies that could distinguish these distinct sources of staining variability. We believe that a laboratory must be aware of both and implement appropriate quality assurance protocols to identify each when it arises. We are now working on extending the potential of this technology to address pre-analytic sources of variability as well. It is our hope that this technology might contribute towards that goal.