Quantitative immunohistochemistry and the use of cellular calibrators for HER2 receptor number determination

Abstract

It is suggested that high error rates in clinical IHC compared to other laboratory disciplines are due to a lack of traceable analytic standards with quantitative controls. Quantitative controls can be achieved by implementing calibrators into IHC workflows, but recently published approaches offer little discussion about what the ideal calibrator should look like. Here, we detail an IHC method that utilizes cellular calibrators to determine the average number of HER2 receptors in tumor cells of breast cancer tissue. Electrochemiluminescent immunoassay and flow cytometry were used to establish nominal HER2 amounts in the cellular calibrators. The quantitated number of HER2 receptors and associated 4B5 scores from commercially sourced breast cancer tumors were compared. The results were discordant, indicating differences in analytical performance. Cells are ideal calibrators for IHC because the antigens they express are structurally and functionally representative of the endogenous analyte being measured. They can also be stained and analyzed identically to tissues, and are easily cultured, expanded, and manipulated. However, novel methods for labeling cells with fixed amounts of proteins are needed. In addition, standards for validation of quantitative IHC methods should be developed and their clinical utility must be demonstrated.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-026-42898-x.

Introduction

Immunohistochemistry (IHC) methods enable the visualization and spatial distribution of antigen expression in complex tissue environments. In oncology, IHC is used to diagnose, subtype, prognosticate, and predict response to a variety of cancers based on biomarker expression patterns in the tumor¹. However, clinical IHC methods have high rates of error compared to other laboratory disciplines². In IHC, errors can arise from the preparation of the specimen, the staining procedure, or the interpretation of the resulting stain³. Results from the NordicQC proficiency testing (PT) study, where participating laboratories analyzed serial specimens that were interpreted by a centralized group of pathologists and biomedical scientists, indicate that analytical components related to the staining procedure are major contributors to error. For example, it was found that approximately 20–30% of results for clinical HER2 IHC methods do not meet clinical standards, primarily due to either too weak or false-negative staining⁴.

Quality control (QC) samples are regularly integrated into clinical laboratory tests to confirm performance of the procedure against expected performance⁵. For clinical IHC methods, positive controls are generally used to monitor protocol sensitivity and calibration of the system, whereas negative controls serve to evaluate the specificity of the test^6,7. However, it is suggested that error persists in clinical IHC due to a lack of traceable analytic standards with quantitative controls^2,8,9. Quantitative controls are critical components to other laboratory disciplines such as ligand binding assays where not only do regulations require that such controls be in place, but there are even specific requirements regarding exactly what the control concentrations should be. For example, to ensure proper analytical sensitivity, the low concentration QC for ligand binding assays must be within 3-fold of the lower limit of quantification (LLOQ) concentration^10,11. While it is recommended that controls containing both low and high levels of analyte also be implemented for IHC methods¹², it is difficult for laboratories to define and compare analytic sensitivities of IHC methods since the control concentrations are typically unknown. This is a potential culprit for the finding that most clinical IHC staining errors are due to either too weak or false-negative staining; laboratories may be unable to verify that low QCs are positioned at antigen amounts close enough to the required analytical sensitivity of the method to reliably detect misperformance.

Traditional IHC scoring systems are typically qualitive (positive or negative), semi-quantitative (the amount of analyte is estimated), or some combination of both¹³. To implement quantitative controls, the IHC readout must be quantitative. This can be accomplished by implementing calibrators. In bioanalysis, calibrators are samples with known amounts of analyte that can be used to interpolate analyte amounts in unknown samples¹⁴. Implementation of calibrators is a straightforward task in many bioanalytical disciplines. For example, ligand binding assay calibrators can be easily prepared by dilution into a biofluid or surrogate matrix to the desired concentrations. In IHC, preparation of calibrators is technically challenging because the target antigen to be measured is bound to cells in a solid tissue sample. Cells or tissues with known amounts of the antigen of interest could be used, or alternatively the antigen of interest could be attached to a solid surface in known amounts. Recently, examples of both approaches have been published. The first involves orthogonal quantification of antigen amounts in established cell lines that are used as calibrators in an IHC workflow^15,16. The second approach involves covalent attachment of antigens to functionalized glass microbeads for the same purpose^17–19. Despite continued interest in quantitative IHC readouts, there has been little discussion regarding how calibrators in IHC should be physically constructed – that is, whether they should be comprised of tissues, cells, or some other form of surrogate cell – to ensure accuracy while remaining practically feasible.

In this paper, we detail an IHC method that utilizes cellular calibrators to quantitate the number of HER2 receptors per tumor cell in breast cancer tissue. To define nominal receptor amounts in the cellular calibrators, HER2 was quantitated using electrochemiluminescent immunoassay (ECLIA) and flow cytometry. Then, the cellular calibrators were implemented into a HER2 IHC method. The method was used to quantitate the average number of HER2 receptors per tumor cell in a tissue microarray (TMA) containing 85 breast cancer patient tumors, and the number of HER2 receptors per cell was compared to the clinical HER2 4B5 result provided with each tumor sample. We chose to study HER2 due to its evolving context of use as a semi-quantitative readout²⁰ as well as increased interest in the potential clinical benefit of histological quantification of HER2 to more accurately stratify patients into the HER-low and -ultralow groups^21,22. However, our primary aim in presenting this work is to summarize the benefits, drawbacks, and challenges of using cells as calibrators in IHC. In addition, we hope to foster discussion about how calibrators in IHC should be physically constructed and how they may be feasibly characterized and implemented to quantify any antigen of interest.

Results

Quantification of HER2 in cellular calibrators

Cell lines corresponding to high and low HER2 expression within the same scoring categories were identified via literature search assigned for the 0, 1 + and 2 + groups (e.g., 1 + Low and 1 + High) and, along with two additional 3 + cell lines and Jurkat negative control, selected as cellular calibrators. Quantitated HER2 values for the cellular calibrators obtained via ECLIA and flow cytometry are provided in Table 1. ECLIA quantification of HER2 in the cellular calibrators as well as Jurkat control lysed on 3 separate occasions is depicted in Fig. 1A and B. The optimized lower limit of quantification (LLOQ) of the ECLIA method was 0.076 pg HER2/µg total protein (1.52 pg/mL; 11 fM). Undetectable levels of HER2 were observed for Jurkat and MDA-MB-468 (0 Low). All other selected cellular calibrators showed quantifiable levels of HER2. Quantitated number of HER2 receptors/cell as determined by flow cytometry are depicted in Fig. 1C. Jurkat was not used in flow cytometry experiments as ECLIA results suggested MDA-MB-468 was an appropriate negative control. All cell lines were quantitated within the kit calibration range (689,000–1,400 receptors/cell) except for AU-565 (3+), which was extrapolated above the top calibrator.

Table 1.

Orthogonal quantification of HER2 in cellular calibrators.

Cell Line	Assigned Nominal Score	ECLIA Result (pg HER2/µg total protein)	Flow Cytometry Result (HER2 receptors/cell)
Jurkat	0	< 0.076	Not analyzed
MDA-MB-468	0 Low	< 0.076	< 1,400
MDA-MB-231	0 High	0.327	2,313
T-47D	1 + Low	1.275	8,203
MDA-MB-175	1 + High	1.580	14,149
ZR-75-1	2 + Low	2.377	22,349
MDA-MB-453	2 + High	4.280	47,050
AU-565	3+	33.985	456,584
BT-474	3+	42.330	277,096

Fig. 1.

Orthogonal quantification of HER2 in cellular calibrators. (A) Quantitation of HER2 by immunoassay in the cellular calibrators lysed on 3 separate occasions. Interpolated cellular calibrator amounts are plotted along the regression. Jurkat (HER2 null control) and MDA-MB-468 (0 Low) are not depicted as HER2 amounts were undetectable for both cell lines. ULOQ and LLOQ are depicted by horizontal dotted lines. Dotted line surrounding the regression is the error. (B) Quantitation of HER2 by immunoassay in the calibrator cell lines lysed on 3 separate occasions. Error bars represent the SD. Jurkat and MDA-MB-468 were below the assay lower limit of quantification (LLOQ) but plotted at the LLOQ for informational purposes. Despite use of MDA-MB-231 as a negative control in HER2 diagnostic assays, our method shows that it has a low but quantifiable amount of HER2. (C) Quantitation of HER2 by flow cytometry in cell lines lysed on 3 separate occasions. Error bars represent the SD. Jurkat was not analyzed as immunoassay data demonstrated that MDA-MB-468 (0 Low) was an appropriate negative control. MDA-MB-468 was below the assay LLOQ but plotted at the LLOQ for informational purposes. (D) Deming Regression of HER2 immunoassay and flow cytometry results. The corrected slope was 1.08, indicating that the methods show strong agreement, however the rank order was reversed in the respective methods for AU-565 and BT-474. Characterization of HER2 amounts in the cell lines allows for nominal concentrations to be assigned to the cell lines for use as calibrators in IHC.

Results for Deming regression are shown in Fig. 1A, B, C and 1D. Correlation was significant (p = 0.005). A corrected slope was calculated such that a slope of 1 shows perfect agreement; the corrected slope was 1.08. Cell lines corresponding to the 0, 1 + and 2 + scoring groups showed rank-order agreement across ECLIA and flow cytometry methods, whereas rank-order was reversed for the 3 + cell lines (AU-565 and BT-474).

IHC Quantitation of HER2 receptors/tumor cell in patient breast cancer tumors

Cellular calibrators were analyzed alongside a TMA containing 85 breast cancer patient tumors, and, using a DIA (digital image analysis) algorithm trained to identify cells and report staining for HER2 on the area of each cell where staining occurred, the number of HER2 receptors per tumor cell was reported. Representative images depicting tumor cell identification and HER2 staining are shown for one of the cellular calibrators, BT-474, and one representative patient tumor sample in Fig. 2. The algorithm can accurately discern between tumor and non-tumor cells by assessing staining for cytokeratin. The algorithm also measures HER2 staining precisely in the area on cells where it is observed.

Fig. 2.

Validation of a machine learning algorithm to identify tumor cells and assess HER2 staining on a per-cell basis in the quantitative HER2 IHC method. Representative images from one of the cellular calibrators, BT-474, and one donor replicate, tumor 42, that expressed quantifiable amounts of HER2 (116,601 HER receptors/cell) are shown to demonstrate the ability of the algorithm to detect individual tumor cells and measure staining in the correct area on each individual cell. The real-time tuning box depicts how cells are identified and where staining occurs by color (selected colors are arbitrary). All identified cells are depicted with a purple nucleus. Cells that are identified as tumor have a yellow outline, and cells that are non-tumor have a red outline. BT-474 is tumor-derived and therefore all cells have a yellow outline. Both tumor and non-tumor cells are depicted in tumor 42. The method measures HER2 only in the yellow area surrounding the tumor cells. The overlay shows that the method is effective at identifying tumor cells and the area on tumor cells that stain for HER2.

Representative staining patterns for the cellular calibrators are shown in Fig. 3A. Although it was intended for each cell line to be embedded in the calibration tissue block at approximately equal density, varying densities resulted. The cellular calibrator regression curve from which HER2 receptor numbers are interpolated is shown in Fig. 3B. MDA-MB-468 (0 Low) was the negative control in the curve, equivalent to a “blank” sample in ligand binding assay calibration. The LLOQ was calculated to be 10,375 HER2 receptors/cell, which is between the number of HER2 receptors established for T-47D (1 + Low; 8,203 receptors/cell) and MDA-MB-175 (1 + High; 14,149 receptors/cell). The upper limit of quantification (ULOQ) was 178,649 receptors/cell, which is well above MDA-MB-453 (2 + High; 47,050 receptors/cell) but below BT-474 and AU-565 (3+; 277,096 and 456,584 receptors/cell, respectively). The calibration curve statistics were R² = 0.910 and SE = 0.247.

Fig. 3.

Staining pattern in the cellular calibrators and resulting quantitative HER2 IHC method calibration curve. (A) Representative images of the cellular calibrators in the quantitative HER2 IHC method. The nominal HER2 receptors/cell as determined by flow cytometry are provided for each cell line. Although cell densities varied, this is unlikely to impact results as staining is assessed on a per-cell basis. An algorithm was developed to identify tumor cells and assess average staining intensity for HER2 on each cell. (B) Calibration curve for the cellular calibrators analyzed in the IHC method. Since IHC is a ligand binding assay, calibration curve fitting principles for ligand binding assays were followed. A 5-parameter fit with 1/y² weighting was used. Error bars represent the SD. MDA-MB-468 is considered the blank and was therefore excluded from the curve fit. R² was 0.910 and SE was 0.247. The number of HER2 receptors per cell in breast cancer tumors was interpolated from the calibration curve. Unlike other techniques that measure analyte concentrations in a biofluid where nominal calibrator concentrations can be easily diluted, it is difficult to achieve evenly spaced calibrator amounts using cell lines. This could impact the reliability of the curve fitting.

Representative images from one of two possible patient tumor replicates, along with the interpolated number of HER2 receptors/tumor cell, are shown in Fig. 4. The full dataset of tumor results, including mean number of HER2 receptors per tumor cell, %CV, vendor-provided HER2 4B5 result, and associated raw Cy5 mean fluorescence intensity (MFI) values, is available in Supplementary Table 1. Most patient tumors (34/49; 69%) had a quantifiable mean number of HER2 receptors per tumor cell. Five patient tumors were below the method LLOQ at all available replicates, five patient tumors were above the ULOQ at all available replicates, and five patient tumors had one replicate quantifiable and the other not quantifiable. Average %CV for patient tumors where both replicates had a valid result was 31%, and in general, %CVs were lower at lower HER2 receptor numbers. The range of quantifiable mean HER2 receptors/cell in patient tumors was 11,326 to 165,507 receptors/cell.

Fig. 4.

Range of quantified HER2 receptor numbers across selected breast cancer TMA samples. Representative images from one of two patient tumor replicates are shown. The average staining intensity across all tumor cells in the sample was used to interpolate the average number of HER2 receptors/cell from the cellular calibration curve. The images are ordered from low (< 10,375 HER2 receptors/cell, the method LLOQ) to high (> 178,649 receptors/cell, the method ULOQ). Tissues that were below the LLOQ or above the ULOQ were ordered by staining intensity. The method LLOQ and ULOQ correspond to categorical HER2 scoring in the 1 + range to in-between 2 + and 3+. Approximately 69% of the total donor replicates analyzed were quantifiable. Some tumors including Tumor 76 and Tumor 178 showed expression that appeared to be heterogenous. The image demonstrates that HER2 expression is continuous and shows what levels of staining intensity are quantifiable using the quantitative HER2 IHC method. Despite extensive optimization, the method was not sensitive enough to quantify HER2 in the HER2-ultralow population but could partially quantify HER2 in the HER2-low population.

Mean HER2 receptors/cell for patient tumors were also grouped by vendor-provided HER2 4B5 score and graphed with the median result in Fig. 5. An inverse correlation was observed; samples with higher number of interpolated HER2 receptors showed lower 4B5 scores. Interestingly, the replicate donor core with the lowest raw MFI (Tumor 198) in the sample set had a 4B5 score of 3 + and the replicate donor core with the highest raw MFI (Tumor 78) had a 4B5 score of 0. Furthermore, among donor cores with a 4B5 score of 3+, there were no > ULOQ results.

Fig. 5.

Individual and median interpolated HER2 receptors per tumor cell in patient breast cancer tumors grouped by 4B5 assay result. Individual interpolated HER2 receptors/tumor cell from the quantitative IHC method are plotted with the median for each group (black bars). The results were discordant; tumors with higher median number of interpolated HER2 receptors were associated with a lower 4B5 score. Among donor cores with a 4B5 score of 3+, there were no > ULOQ results, and most tumors with a 0 4B5 result showed quantifiable HER2 receptor numbers. While such discordance is notable, the root cause is likely analytical differences between the 4B5 method and our quantitative method. There is significant interest in quantitative IHC readouts, however guidelines for validation of such methods must be implemented and it must be shown that they have clinical relevance.

Discussion

Establishing the nominal amount of HER2 in the cellular calibrators is critically important for accurate IHC receptor number interpolation. Our results showed that both ECLIA and flow cytometry are feasible approaches, although the methods differed in their quantitation of HER2 in the 3 + cell lines, AU-565 and BT-474. Aside from being separate technologies, ECLIA and flow cytometry utilized different primary antibodies (29D8 in ECLIA versus Mab1129 in flow cytometry) and measure HER2 in a different cellular setting (lysed cells in ECLIA versus live, unfixed cells in flow cytometry). We suspect that variable lysing efficiency in the 3 + cell lines may explain the difference, since the two approaches that measure HER2 directly on cells, flow cytometry and IHC, achieved similar results using different antibodies (29D8 was used for IHC). ECLIA was more sensitive than both flow cytometry and IHC, and this allowed accurate quantitation of MDA-MB-231. Notably, MDA-MB-231 is the negative control in several HER2 diagnostic IHC methods, where it is expected that no staining should occur. Our results suggest that if the IHC method is sensitive enough, MDA-MB-231 should stain for HER2, and that MDA-MB-468 may be a more appropriate negative control for clinical HER2 methods. Sensitivity of the bioanalytical approach is an important consideration especially in the context HER2-ultralow which corresponds to the range between 0 and 1+.

IHC is, in essence, a ligand binding assay that takes place on a tissue. Therefore, we applied ligand binding assay curve fitting and quantification limit calculation principles to the quantitative HER2 IHC method. This included the use of a 5-Paramter Logistic calibration curve with 1/y² weighting. The quantifiable range of the method was 10,375 to 178,649 HER2 receptors/cell, which corresponds to categorical HER2 scoring in the 1 + range to in-between 2 + and 3+. This quantifiable range spans less than 2 orders of magnitude, which is narrow but expected for IHC¹⁸. Despite utilizing tyramide amplification and performing extensive optimization, our quantitative IHC method was not sensitive enough to quantify HER2 in the HER2-ultralow population but could partially quantify HER2 in the HER2-low population. While we cannot conclude with quantitative certainty the level of sensitivity offered by HER2 diagnostic IHC methods since they do not contain calibrators, it was recently shown that the 4B5 method may be unable to reliably discern HER2-ultralow from HER2-null cases²³. We therefore expect that the sensitivity of our quantitative method is approximately similar to other diagnostic methods in use. It is also worth noting that the cellular calibrators varied in density, as shown in Fig. 3, likely because the cell numbers prior to embedding and processing were estimated from confluency. This should have little bearing on the result since HER2 staining was measured on each individual cell. However, staining patterns may differ when cells are tightly packed such as in a tumor sample.

We compared the median number of HER2 receptors/cell for each patient tumor to the vendor-provided 4B5 assay score and found that patient tumors with higher interpolated HER2 receptors/cell had lower 4B5 scores. These results should be interpreted with caution for several reasons. First, our quantitative method is not fully validated, and our results were obtained from a single run where the method LLOQ and ULOQ were estimated formulaically, rather than empirically. In addition, our method showed an average %CV of 31% across patient tumors with two valid replicates, although there are currently no industry standards for accuracy and precision for quantitative IHC methods and such results could be due to tumor heterogeneity. Analytical differences between the 4B5 method and our quantitative method are likely the root cause of the discordant results. It is well known that different biomarker methods are expected to show varying levels of accuracy, precision, specificity, and sensitivity^24,25, Notably, HER2 clone 4B5 is known to cross-react with HER4 and several other proteins^26,27. We thus did not consider use of 4B5 and instead selected 29D8 in our method which outperformed two other antibody clones tested. Tumor heterogeneity could also be a contributing factor to the discordant results, especially since our tumor cores were 1 mm, which is smaller than the recommended 5–10 mm²⁸.

Industry standards for quantitative biomarker methods recommend choosing a calibrator that is structurally and functionally representative of the endogenous analyte found in clinical samples²⁹. In IHC, the target antigen is bound to a cell, and thus the ideal calibrator should also be the target antigen bound to a cell. This may include monocultured cells, such as cell lines, or tissues containing cells that express the antigen of interest. Practically, the most feasible approach is to use cell lines since they can be easily cultured, expanded, and manipulated. As with other clinical laboratory disciplines, however, it is critical that QCs be implemented that match the tissue being analyzed. Cells as calibrators also allow the flexibility to apply either chromogenic or fluorescent staining techniques since they can be stained the exact same way as tissues, for example with Hematoxylin and Eosin or DAPI to identify subcellular structure. This overcomes one limitation with the use of recombinant antigen that is attached to surrogate cells.

However, there are several limitations of the approach presented here. First, the cellular calibrators were not prepared from scratch; they were selected from a plethora of already-established cell lines. This strategy is not applicable to antigens that are not expressed in cells at adequate levels or for which established cell lines do not exist. In addition, reliance on existing cell lines resulted in unequal spacing across the calibration curve, which is suboptimal for calibration curve fitting. This could lead to inconsistent accuracy and precision of the method from run-to-run. Future studies could improve upon this by developing novel methods for labeling cells with fixed amounts of protein antigens, which could include antigens that are not endogenous (for example, drugs). We also developed orthogonal bioanalytical methods to characterize HER2 amount in the cellular calibrators, which was time consuming and resource intensive. Antigen agnostic bioanalytical approaches to measure cells labeled with protein antigens would significantly streamline the approach.

Conclusions

In this study, we characterized and implemented cellular calibrators into an IHC method workflow that quantitated the number of HER2 receptors in breast cancer tissue tumor cells. When quantitation in IHC is desired, cells lines are ideal calibrators because the antigens they express are structurally and functionally representative of the endogenous analyte found in tissue samples. Cells can be stained the exact same way as tissues, and are readily cultured, manipulated, and expanded. However, the approach used to define nominal cellular calibrator concentrations matters because different platforms and detection systems vary in their accuracy, specificity, and sensitivity. The use of cellular calibrators for IHC would be further streamlined by novel methods for labeling cells with fixed amounts of proteins that can be measured using an antigen agnostic approach.

We detailed how existing cell lines and orthogonal bioanalytical approaches may be used to quantitate receptor number tumors. While it is interesting that the average number of HER2 receptors in tumor cells of each breast cancer sample was discordant with the provided 4B5 scores, it is important to caution that our method is not validated analytically or clinically. Standards for validation of quantitative IHC methods must be developed. Since IHC is a ligand binding assay that takes place on a tissue, quantitative ligand binding assay validation standards for biomarkers may be a useful starting point for such efforts. In addition, for quantitative IHC to be of use, it must have clinical relevance. With this work, we do not suggest that our quantitative HER2 IHC method is superior to current clinical methods in use; we simply present it as an example for how quantification can be achieved. Efforts to implement quantitative IHC readouts should include a rigorous comparison to current qualitative and/or semi-quantitative methods and must demonstrate that quantitation aids in the clinical utility to diagnose, subtype, prognosticate, or predict patient responses.

Materials and methods

Cell Culture

Jurkat, MDA-MB-468, MDA-MB-231, MDA-MB-175, MDA-MB-453, and BT-474 were sourced from GSK (Collegeville, PA, USA). T-47D, ZR-75-1, and AU-565 were purchased from American Type Culture Collection (ATCC; Manassas, VA, USA). All cell lines were grown in RPMI-1640 (ThermoFisher, Waltham, MA, USA; catalog 11875085) supplemented with 10% heat-inactivated FBS (ThermoFisher; catalog 10082147) and 1x Penicillin-Streptomycin-Glutamine (ThermoFisher; catalog 10378016) at 37 °C and 5% CO₂. Cell lines were passaged no more than 8 times prior to use in all experiments. When cells reached approximately 70% confluency, PBS-based enzyme-free cell dissociation buffer (ThermoFisher; catalog 13151014) was used for dissociation prior to passage or experimental use.

HER2 ECLIA and analysis

Cells were lysed for 30 min with 1 mL M-PER Mammalian Protein Extraction Reagent (ThermoFisher; catalog 78501) per 5 × 10⁶ cells. Lysates were subsequently aliquoted and stored at −70 °C until further use. BCA assay (ThermoFisher; catalog 23225) was used to determine total protein concentration of respective cell lysates according to kit instructions. Lysates were prepared on 3 independent occasions.

Anti-HER2 rabbit mAb clone 29D8 (Cell Signaling, Danvers, MA, USA; catalog 2165 S) was diluted to 4 µg/mL in 1x PBS and adsorbed onto a standard bind plate (MSD; catalog L15XA). The following day, the plate was washed and then blocked with Diluent 43 (MSD; catalog R50AG) for 1 h. The calibrator, full-length HER2 protein from Origene (Rockville, MD, USA; Catalog TP710032), was prepared from 10,000 to 0.170 pg/mL in diluent 43 along with a blank (0 pg/mL). Lysed samples were diluted to 20 µg/mL total protein for normalization. The blocked plate was washed, and calibrators and samples (1 µg total protein/well) were added to the plate in duplicate and incubated for 2 h. Following incubation, the plate was washed again, and anti-HER2 goat pAb from R&D Systems (Minneapolis, MN, USA; Catalog AF1129) was diluted to 2 µg/mL and added to the plate for 1 h. The plate was washed a final time and then 2x read buffer (MSD; catalog R92TC) was added to the plate. The plate was immediately analyzed on a Sector S600 reader where electrochemiluminescent (ECL) signal was obtained. Lysates prepared on 3 separate occasions were analyzed once.

Results from the three lysate preparations were pooled for data analysis using MSD Workbench software. Calibrators were fitted using a 4-PL curve with 1/y² weighting. The assay upper limit of quantification (ULOQ) and lower limit of quantification (LLOQ) were the highest and lowest calibrator points with acceptable accuracy and precision (%CV and %RE ≤ 25%). Sample results were interpolated in pg/mL and then converted to pg HER2/µg total protein and fM.

HER2 flow cytometry and analysis

QIFIKIT Quantitative Analysis Kit from Agilent (Santa Clara, CA, USA; catalog K0078), which uses a series of beads coated with known quantities of mouse mAb (689,000–1,400 receptors/cell) as calibrators, was used to quantify the number of HER2 receptors per cell in each cell line using a mouse anti-HER2 mAb (R&D Systems, catalog Mab1129). Titrations of the primary and secondary antibodies were used to optimize the sensitivity of the method. A mouse IgG isotype control (ThermoFisher, catalog 31903) was used to confirm specificity of the method. 50,000 cells from each cell line were plated onto a polypropylene dilution plate and washed 3x with 1% BSA in PBS w/10% FBS (buffer). Mouse anti-HER2 antibody was diluted to 10 µg/mL in buffer and incubated with the cells for 30 min. Cells were washed with buffer and goat anti-mouse FITC pAb (ThermoFisher; catalog 31569) was diluted to 5 µg/mL in buffer and incubated with the cells for 30 min. The cells were washed a final time re-suspended to a density of approximately 250,000 cells/mL in 1x PBS w/10% FBS for analysis on BD Biosciences (Franklin Lakes, NJ, USA) FACSCanto flow cytometer.

According to the kit instructions, setup beads were used to establish the data acquisition condition for the calibration beads, which was optimized at FSC 30 V, SSC 275 V, and FITC 350 V. The acquisition condition for cell lines was determined independently of the beads and optimized to be FSC 30 V, SSC 320 V, and FITC 350 V. Data from 10,000 beads or cells were acquired from each sample. Three independent experiments were run on three different days, with each run using cell lines that were freshly thawed from liquid nitrogen storage.

FlowJo (v10.8.1) was used for data analysis. The FITC MFI of the calibration beads minus the MFI of the blank beads without HER2 was plotted against the number of HER2 molecules per bead to generate a calibration curve using GraphPad Prism (v10). The number of HER2 receptors per cell for each cell line was interpolated from the calibration curve. Data from three independent experiments were used to calculate the average number of HER2 receptors for each cell line.

Correlation between ECLIA and flow cytometry

Deming regression was used to compare quantification of HER2 by ECLIA and flow cytometry since there is statistical uncertainty in values obtained from both methods. Average standard deviations across measurements for each method were inputs for the regression. The slope and standard error (SE) of the slope and y-intercept were reported. A corrected slope was also reported by calculating a correction factor for each cell line (flow cytometry result divided by ECLIA result for each cell line) and dividing the calculated slope by the mean corrected result across all cell lines.

Quantitative HER2 IHC method

The number of HER2 receptors per tumor cell in breast cancer tumors was determined by using the known amount of HER2 receptors in the cellular calibrators, as quantitated by flow cytometry, which were embedded into a tissue block, sectioned onto microscope slides, and then analyzed alongside tumor samples as calibrators.

Cells were grown in dedicated T175 flasks until approximately 70% confluent, which corresponds to around 25 million cells. After dissociation the cells were fixed in 10% NBF (ThermoFisher; Catalog 23–730-588) for 1 h at room temperature. The cells were washed with PBS, embedded in 100 µL HistoGel (ThermoFisher; Catalog 22–110-678), placed on ice to solidify, and fixed again with 10% NBF overnight. The following morning, the pellets were placed in a tissue cassette and processed into a paraffin wax block. The block was cut into 4 μm sections which were mounted onto AutoFrost microscope slides (Cancer Diagnostics, Durham, NC, USA; catalog EMP00W3-BX). The slides were stored at room temperature until use.

Staining was performed using Roche Ventana (Tucson, AZ, USA) DISCOVERY ULTRA automated staining system. Three replicates of cellular calibrators separated across three slides were analyzed alongside a TMA containing breast cancer cores from approximately 90 patients (Tristar Technology Group, Rockville, MD, USA; catalog TA1621) in one run. The tissue samples were sourced ethically, and their research use was in accord with the terms of the informed consents under an ethics committee approved protocol (Istituto Oncologico del Mediterranean, reference no. 1313/CE). The method was optimized to quantify HER2 on tumor cells using the Ventana Discovery HQ amplification system. Slides were deparaffinized over 3 cycles at 69 °C for 8 min each. Epitope retrieval was performed for 24 min at 94 °C. One drop of Discovery Inhibitor was applied for 8 min to block endogenous peroxidase activity. Primary HER2 antibody clone 29D8 was applied at 8 µg/mL for 32 min at 36 °C, followed by anti-rabbit HQ for 32 min at 36 °C, anti-HQ HRP for 28 min, Cy5-tyramide for 4 min, and then Cy5-H₂0₂ for 28 min. Discovery inhibitor was reapplied to each slide to quench previously bound HRP. Primary anti-pan keratin was then applied (46.3 µg/mL) for 16 min at 36 °C, followed by anti-mouse HQ for 16 min, anti-HQ-HRP for 16 min, FAM-tyramide for 4 min, and then FAM-H₂0₂ for 16 min. All slides were counterstained with spectral DAPI (Akoya Biosciences, Marlborough, MA, USA) for 8 min. Following completion of the method, slides were washed and baked for 5 min at 60 °C. ProLong Gold Antifade Mountant (ThermoFisher; catalog P36930) was applied to each slide which was coverslipped and allowed to cure overnight prior to imaging.

Images were acquired using the Vectra Polaris imaging system (Akoya Biosciences) at 20X magnification. HER2 (Cy5), pan-keratin (FAM), and nuclei (DAPI) were imaged using the Opal 690, Opal 520, and DAPI filters, respectively, with exposure times of 12, 1, and 3 ms, respectively.

Images were uploaded to HALO and analyzed using HALO AI v4.0 (Indica Labs, Albuquerque, NM, USA). DIA was implemented as it is standard practice in our lab for IHC methods utilizing fluorophores. A nuclei segmentation classifier was trained to identify cell nuclei using a resolution of 0.25 μm/pixel. A minimum threshold for nuclei size was not applied. The classifier was implemented into an analysis protocol with the following parameters: Minimum Nuclear Intensity = 0, Maximum Image Brightness = 0.4, Nuclear Segmentation Aggressiveness = 0, Nuclear Size = 20,500, Minimum Nuclear Roundness = 0.272. Due to difficulty in training HALO to properly identify cell membranes, HER2 (Cy5) and pan-keratin (FAM) staining was measured in the cytoplasm compartment, the radius of which was assigned as 2.5 μm from the identified nucleus. The accuracy of the algorithm measure HER2 and pan-keratin staining where staining occurred was validated visually. Cell size was bound between 70 and 500 µm². MFI in the cytoplasm compartment for HER2 and pan-keratin was reported on a per-cell basis for either tumor cells or non-tumor cells. For tumor cell identification, a cutoff of 20 MFI for pan-keratin was applied. Tumor cores that were folded and/or creased were omitted from the analysis. In addition, tumor cores that contained less than 100 tumor cells were omitted from analysis.

MFI was reported on a per-cell basis for cellular calibrators and tumor cells in the TMA breast cancer cores. A calibration curve using HER2 receptors/cell as determined by flow cytometry as the nominal receptor amount for each cell line was regressed against HER2 MFI response for the cellular calibrators, and mean MFI from tumor cells in each core was used interpolate the number of HER2 receptors in each sample.

Since IHC is a ligand binding assay, calibration curve fitting principles for ligand binding assays were followed where feasible. GraphPad Prism v10 was used to regress the calibration curve and interpolate unknowns. A 5-PL, 1/y² was determined to fit the data best as determined by R². The regression was conducted with cellular calibrators in triplicate and SE was also reported. Since one run was performed, LLOQ was determined analytically by calculating the mean MFI of the blank responses + 10 SD (MDA-MB-468 is considered the blank). The blank was not included in the regression fit. Since there was inconsistency in measured HER2 content in the two amplified cell lines across ECLIA and flow cytometry (see results section), the ULOQ was interpolated from the upper asymptote of the regression.

Summary of HER2 receptor number across tumor samples

The breast cancer tumor TMA was purchased from Tristar Technology Group (Rockville, MD, USA; catalog TA1621). Although the tissue block from which the TMA was cut contained 180 cores (2 cores each from 90 patients), results from only 85 replicate cores were suitable for analysis. Most cores were simply not present in the TMA sample when it was received; others were either folded and/or creased or contained < 100 tumor cells and thus were omitted from the analysis.

Mean HER2 receptor number per tumor cell was reported for replicate breast cancer tumors. Where suitable results were obtained from both replicates, % CV was reported, and the vendor-provided 4B5 staining result was also reported for each patient tumor. According to the vendor, the 4B5 result was obtained by a CAP/CLIA accredited laboratory. Where one replicate for a patient tumor was quantifiable and one replicate was outside of the quantifiable method range, the LLOQ or ULOQ value was substituted to facilitate calculation of means. Mean HER2 receptors per tumor cell for each patient tumor were also grouped by HER2 4B5 result with median. An image of stained tumor samples inclusive of one image from each patient was compiled along with the interpolated number of HER2 receptors per tumor cell to demonstrate the continuous range of HER2 expression and dynamic range of the method.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

K.M. and B.C. conceptualized the research. K.M. developed all bioanalytical methods, acquired all data, analyzed all data, and drafted the manuscript. B.C. supervised the project and contributed to the writing and editing of the manuscript.

Funding

GSK provided funding for this study. K.M. was employed by GSK when the work was completed and is a shareholder of GSK. B.C. declares no potential conflicts of interest.

Data availability

The data that support the findings of this study are provided with the supplementary material.

Declarations

Competing interests

The authors declare no competing interests.