A Comparison Between Immunohistochemistry and mRNA Expression to Identify Human Epidermal Growth Factor Receptor 2–Low Breast Cancer Open Access

In this study, we aim to assess the relationship between HER2 expression detected by IHC (majority score after evaluation by 16 pathologists) and mRNA using the MammaTyper (Cerca Biotech distributed by Sysmex Europe SE) quantitative reverse transcription–polymerase chain reaction (RT-qPCR) kit.

We selected all BC needle biopsies (n = 105) from our previously published interobserver study of nonoverexpressing HER2 BCs.¹⁸  This was a consecutive series of patients diagnosed in 2019. In this study, each case underwent digital HER2 IHC evaluation by 16 specialized pathologists. All slides were scanned with the Nanozoomer 2.0-HT (Hamamatsu Photonics) and uploaded to Slide Score B.V. (version 1.2-2022-05-24T15:37:11) (Netherlands Cancer Institute, Amsterdam, the Netherlands), which allowed zooming to a high magnification (objective, ×40). Scoring was performed manually by the pathologists. HER2 IHC was performed with the VENTANA HER2 rmab clone 4B5, 790-4493 (laboratory-modified protocol settings) on an automatic immunostainer (Ventana BenchMark Ultra, Roche), according to the manufacturer’s instructions. The following criteria were used for the IHC scoring: IHC 0 (no staining), IHC >0, <1⁺ (incomplete membrane staining in ≤10% of tumor cells), IHC 1⁺ (incomplete membrane staining in >10% of tumor cells), and IHC 2⁺ (weak to moderate complete membrane staining in >10% of tumor cells). Our HER2 IHC staining is evaluated by the Nordic Immunohistochemical Quality Control (NordiQC) 2 times per year, with an optimal result.

A high majority agreement was reached when 75% or more of pathologists (12 of 16) agreed on 1 IHC score, and low agreement occurred when fewer than 75% but more than 50% of pathologists agreed. In addition, fluorescence in situ hybridization (FISH) was performed with the BenchMark Ultra. For the detection of HER2, the ZytoLight SPEC ERBB2/CEN 17 Dual Color Probe (Zytovision) was used according to the manufacturer’s protocol. Signal numbers for the chromosomal region 17q12-q21.1 harboring the HER2 gene (labeled with SPEC ERBB2) and the alpha satellite centromeric region of chromosome 17 (CEP17; labeled with CEN 17, ZytoLight) were counted in at least 30 invasive tumor cells, and the ratio of HER2 to CEP17 signal numbers was calculated. The assessment was done by one observer and a second observer checked and approved the interpretation of the tests according to the ASCO/CAP 2018 guideline.⁶  We reclassified the tumors in the following categories based on the previous tests: HER2-0 (IHC 0), HER2-ultralow (IHC >0, <1⁺), HER2-low (IHC 1⁺ or 2⁺ with negative FISH result).

For the current study, cells from invasive cancer were microdissected to ensure a reliable comparison with the HER2 IHC result, which is also only scored in invasive tumor cells. Prior to microdissection, the areas of invasive tumor were marked by a breast pathologist using the last 4-µm-thick hematoxylin-eosin–stained slides of 5 sequential slides. The remaining 4 slides of formalin-fixed paraffin-embedded tissue were dewaxed and rehydrated. Following hematoxylin staining, invasive cells were miscrodissected manually by using a sterile scalpel under stereomicroscope (Zeiss, Oberkochen, Germany) and stored into RNAse/DNAse-free tubes containing RNALater (Thermo Fisher). RNA was isolated with the RNeasy FFPE kit (Qiagen) according to manufacturer’s instructions. MammaTyper RT-qPCR was used for quantitative evaluation of HER2. Cutoff values of the MammaTyper assay according to manufacturer’s instructions were used as reference for interpretation of mRNA levels.^24–27  These corresponded to 3 groups based on ΔΔCq cutoff values: HER2 0/ultralow (<34–38 ΔΔCq), HER2-low (≥38–40.9 ΔΔCq), and HER2-positive (≥40.9 ΔΔCq). Since completing this study the cutoff for HER2 positivity has been lowered to 40.2 ΔΔCq by the manufacturer. We compared the mRNA expression levels with the IHC scores with the highest agreement. Post hoc observational analysis of discordant cases was performed by 1 breast pathologist.

For tumors with discordant HER2 IHC scores, we used the IHC score most frequently scored by the pathologists in the final analysis. When there was no majority score, the highest score was used for analysis. The χ² test was used to compare categorical variables between IHC categories and MammaTyper categories. The Shapiro-Wilk test was used to check all variables for normal distribution. Analysis of the association between the IHC scores and mRNA levels was performed by applying the Kruskal-Wallis and Mann-Whitney U tests and expressed as mean rank, as these data were not normally distributed. Linear regression analysis was conducted to predict the IHC-based HER2 scores by FISH and MammaTyper. A P value of <.05 was considered significant, except for the post hoc Mann-Whitney U tests following the Kruskal-Wallis tests, where a value of P < .016 and P < .025 was used according to Bonferroni correction for multiple testing. Analyses and figure generation were performed in SPSS (version 28.0.1 [released 2021], IBM Corp).

This study was approved by the Medical Ethics Committee of Erasmus MC, Rotterdam, the Netherlands (MEC-2023-0700). The Medical Research Involving Human Subjects Act does not apply to this work; consequently there was no need for an informed consent. The study was performed in accordance with the Declaration of Helsinki.

The methodology for case selection and analysis with MammaTyper is summarized in Figure 1. From the original 105 cases used in our previous interobserver study, 88 had enough tissue for mRNA extraction and valid RT-qPCR results. From these 88 cases, 11 (12.5%) were scored as IHC 0, 6 (6.8%) as IHC ultralow, 45 (51.1%) as IHC 1⁺, and 26 (29.5%) as IHC 2⁺ by most pathologists. All cases were nonamplified according to FISH. After reclassification, 17 cases (19%) were HER2 0/ultralow and 71 (81%) were HER2-low. In 2 cases, the number of pathologists’ votes across IHC categories was equal. The first case without a majority diagnosis was 0/ultralow versus 1⁺ and the second was 1⁺ versus 2⁺. The agreement between IHC scores and mRNA expression per case is illustrated in Figure 2.

Overall, the agreement between the HER2 groups (HER2 0/ultralow and HER2-low) according to IHC and MammaTyper was 84% (n = 74 of 88). From the 14 discordant cases, 10 corresponded to HER2 0/ultralow by IHC and HER2-low by RT-qPCR, 2 cases were HER2-low by IHC and HER2 0/-ultralow by RT-qPCR, and 2 were HER2-low by IHC and HER2-positive by RT-qPCR. In other words, 10 of 17 HER2 0/ultralow cases by IHC (58.8%) disagreed with MammaTyper, and 4 of 71 HER2-low cases by IHC (5.6%) disagreed with MammaTyper. Those 2 cases that were classified as HER2-positive by mRNA were scored as 2⁺ by IHC with a high agreement between pathologists, as illustrated in Figure 2. There were no IHC 3⁺ scores. Examples of cases with a disagreement between HER2 IHC and mRNA are shown in Figure 3, A and B.

Post hoc analysis of cases with disagreement between IHC and mRNA quantification did not show any recognizable clinical or pathologic characteristics. Some of the discrepant cases showed unspecific nonmembranous HER2 staining, but this was also seen in some of the concordant cases, so this was not an obvious explanation for discordance between IHC and mRNA expression. We also evaluated if there was an association between the level of pathologist agreement (<75% versus ≥75% agreement) and the discordance between IHC and mRNA expression. From the 14 cases that were discordant between IHC and MammaTyper, 3 had less than 75% agreement among pathologists and 11 had an agreement of 75% or more, so the discrepant cases could not be explained by a lower pathologist agreement. Furthermore, a significant difference in mean rank levels of HER2 was observed by comparing the continuous variable of mRNA expression to the IHC scores with HER2 0 and ultralow merged together (P < .001). After Bonferroni correction (P < .025), a significant difference in mean rank was also found between groups (IHC 0/ultralow versus IHC 1⁺: P < .001; IHC 1⁺ versus 2⁺: P = .018) (Figure 4, A). This analysis was then performed with the groups IHC 0 and ultralow separated (Figure 4, B). After Bonferroni correction (P < .016), no significant difference in mRNA expression was found between IHC groups (IHC 0 versus ultralow: P = .54; ultralow versus 1⁺: P = .02; 1⁺ versus 2⁺: P = .018). However, when comparing the mRNA expression of IHC 0 versus 1⁺, a significant difference was achieved (P < .001).

Dual-probe FISH was performed in all cases and no amplification was observed. The median and range of HER2 copy number and the HER2 to CEP17 ratio was 1.8 (1.06–4.23) and 1.59 (0.432–4.188), respectively. A significant difference in median levels of HER2 copies was achieved by comparing the continuous variable of HER2 copy number to the IHC scores with HER2 0 and ultralow merged together (P < .001). After Bonferroni correction, a significant difference in mean rank was found between IHC 1⁺ and IHC 2⁺ (P = .006), but not between IHC 0/ultralow and 1⁺ (P = .06) (Figure 5, A). This analysis was then performed with the groups IHC 0 and ultralow separated (Figure 5, B). A significant difference was present only between IHC 0 versus 2⁺ (P = .004) and IHC 1⁺ versus 2⁺ (P = .006).

Furthermore, a linear regression analysis was performed to evaluate the agreement between HER2 copy number detected by FISH and the HER2 mRNA expression. There was a moderate association between these variables (r = .54, P = .01) (Figure 6).

A multiple linear regression analysis was conducted to predict the HER2 scores by IHC in nonamplified tumors, based on FISH and MammaTyper. Overall, the independent variables significantly predicted HER2 score by IHC (F_2,85 = 22.54, P < .001), which indicates that both ISH and mRNA expression are significantly related to the IHC result. Additionally, coefficients were further assessed to ascertain which variable was stronger to predict IHC scores. This showed that HER2 copy number assessed by FISH had a lower impact on predicting IHC (B = .23, t = 2.52, P = .013) than mRNA expression evaluated by MammaTyper (B = .48, t = 5.29, P = .01).

Since the recent introduction of T-DXd for treatment of metastatic HER2-low BC, the search for more precise diagnostic methods to detect HER2-low BC has become very relevant. The current diagnostic techniques recommended by international guidelines for the detection of patients who could benefit from regular HER2-targeted therapy have proved less accurate to detect HER2-low tumors.^6,14,18  The main reason for this is that IHC and FISH are calibrated to discriminate between HER2-amplified versus nonamplified tumors, while the difference between HER2 0 and HER2-low may be so limited that it cannot be detected accurately by these conventional methods.¹⁵ 

In the present study, we assessed the agreement of HER2 status between IHC by a majority of pathologists’ votes and mRNA expression in nonamplified HER2 BC cases. Based on the cutoff values provided by MammaTyper, a high agreement (84%) was found between IHC and mRNA expression. Most of the cases that were discordant corresponded to tumors that were scored by IHC as HER2 0/ultralow and as HER2-low by mRNA expression, meaning that more than half of the HER2 IHC 0-ultralow cases (10 of 17, 58.8%) had a mRNA expression comparable to that of HER2-low tumors. Furthermore, we demonstrated that there was a significant difference in mean rank mRNA expression between the IHC groups when HER2 0 and ultralow were grouped together. However, this difference was negligible once the groups IHC 0 and ultralow were separated. Similar results were observed in 2 previous studies, where mRNA expression showed substantial overlap between IHC 0 groups and IHC 1⁺ regardless of a significant difference in mean rank expression.^15,28  This suggests that HER2 mRNA levels might be indistinguishable across these IHC groups of nonamplified tumors. On the other hand, some authors have reported a significant difference in mRNA expression between IHC 0 versus 1⁺, and ultralow versus 1⁺, but not a significant difference between IHC 0 versus ultralow.^21,29,30  These results indicate that the IHC assay has limitations regarding the detection of low levels of HER2 expression, since the HER2 0/ultralow group seems to be a heterogeneous cluster with certain cases expressing very low HER2 mRNA levels, and other cases having a higher expression comparable to that of HER2-low tumors. An in vitro study established that cells containing more than 20 000, 100 000, 500 000, and 2 300 000 receptors corresponded to IHC 0, IHC ultralow, IHC 1⁺ or 2⁺, and IHC 3⁺.³¹  The slight difference in receptor number between IHC 0 and ultralow could explain the difficulty to distinguish IHC 0 from ultralow with the current diagnostic methods.

Another factor to consider is the HER2 antibody calibration. In this study, the HER2 antibody PATHWAY 4B5 (Roche Diagnostics) was used, which is the one used in the DESTINY-Breast trials. However, a recent study by Rüschoff et al³²  reported that PATHWAY 4B5 (Roche Diagnostics) may have a lower sensitivity to detect HER2-low tumors when compared to the HercepTest mAb pharmDx antibody (Dako Omnis). Thus, the results of this study relate to the antibody used and are not generalizable to other HER2 antibodies.

In addition, we compared the ability of mRNA expression and FISH to predict IHC scores. mRNA expression assessed by MammaTyper showed a higher association to IHC than FISH. In our previous interobserver study, we reported that FISH is not a useful tool to distinguish among HER2-nonamplified subgroups.¹⁸  In this current study, there were 10 cases (of 14 discordant cases) that MammaTyper classified as HER2-low, while IHC classified them as HER2 0/ultralow. This overestimation could be attributed to the perils of IHC itself. As previously noted by our group, numerous tumors with significant discordance among pathologists contain regions with nonspecific HER2 staining, adding complexity to the scoring process.¹⁸  It may be that cells exhibiting these nonspecific staining patterns have some HER2 expression, but owing to the influence of preanalytical factors this is not reflected correctly in the slide. On the other hand, another reason for this phenomenon could be attributed to the nature of mRNA quantification by RT-qPCR, which assesses HER2 expression on a continuous scale, rather than a categorical classification. This distinction might elucidate why the measurement of mRNA expression levels could surpass the accuracy of IHC in certain cases.^15,28 

An important factor to consider is the loss of spatial information in a PCR test.³³  The sample dilution does not consider intratumor heterogeneity and other cell types from the tumor ecosystem.³³  For instance, in a study by Acs et al,³⁴  it was observed that an increase in stromal cellularity and/or associated inflammatory cells could contribute to an increase of false negatives and false-positive results in RNA quantification. However, a study by Viale et al³⁵  argued that discordances between RNA and HER2 IHC could not be explained by intratumoral heterogeneity. Another factor that has been related to higher HER2 mRNA levels is the presence of ductal carcinoma in situ (DCIS) with higher IHC expression levels than the invasive component.³³  In a MammaTyper study by Hartmann et al,³⁶  which included both HER2-amplified and nonamplified BC cases, the mean rank difference in RNA expression between whole sections containing up to 60% DCIS and specimens after microdissection was only 0.16 cycles for HER2 (0.00 to 0.79), although the authors noted that there was a tendency for higher expression in DCIS. These results are similar to other publications.^37,38  However, the differences in HER2 expression between IHC 0/ultralow and 1⁺ are so small that the impact of these variables in mRNA quantification should be considered. In our present study, microdissection was performed to reduce the influence of all these factors and to optimize the association with the IHC score, which was also limited to the invasive component only.

Until now, the clinical indication for treatment with T-DXd was restricted to HER2-low cases, so more precise methods to detect low levels of HER2 expression with clinical relevance are required. However, the optimal lower threshold is unknown. Recently, the effect of T-DXd was tested in the DESTINY-Breast06 trial, which included 153 patients with HER2-ultralow metastatic BC. Here T-DXd reduced the risk of disease progression or death by 22% as compared to chemotherapy with a median progression-free survival of 13.2 months versus 8.3 months, respectively.³⁹  These results in the HER2-ultralow group are consistent with the HER2-low group, so this study showed that T-DXd could also be an effective treatment option for patients with HER2-ultralow BC. Results from the DESTINY-Breast15, which includes patients with HER2-0 metastatic BC (with and without hormone receptor positivity) may further clarify the clinical relevant lower threshold of HER2 expresssion.^39,40  Currently, data from the DAISY trial (including a total number of 177 patients) indicate that among the 37 patients categorized as having an IHC 0 score and treated with T-DXd, 11 had an objective response.²³  Additionally, central review of 31 of these 37 cases revealed some level of HER2 expression by IHC in 15 samples (8 were classified as IHC ultralow and 7 as IHC 1⁺) and no detectable IHC HER2 expression in 16 cases. Among these, 6 of 15 patients with detectable HER2 expression and 4 of 16 patients without detectable IHC HER2 expression had a confirmed objective response.²³  Additional large-scale studies of HER2 nonamplified BCs (including HER2 0/HER2 ultralow cases) with clinical outcome are needed to determine the lower level of clinically relevant HER2 mRNA expression. The performance of several diagnostic tests (eg, IHC and mRNA expression) should be linked to these studies, which enables the comparison of their performance in relation to therapy response. In other words, these future studies should answer the following question: Are HER2 mRNA levels better predictors for therapy response than HER2 protein expression levels? Based on its dynamic range, reflecting a biological continuum, mRNA expression could have additional value compared to the categorical IHC score in the future. However, until these tests are compared with one another and linked to clinical response to T-DXd, we believe that IHC, regardless of its limitations, is still the most practical and validated test to detect HER2-low in routine clinical practice now.

The present study offers a comprehensive examination of mRNA levels across HER2 0, ultralow, and low BCs. A notable aspect of this study is the central review of IHC scores by a considerable number of expert pathologists. Another strength of this study is the use of microdissection before mRNA extraction, minimizing the potential impact of other cell types on mRNA quantification. However, this study also has certain limitations. The most important limitation is the number of patients, mainly those cases classified as IHC 0 and ultralow. Another limitation is the lack of knowledge of clinical response to T-DXd.

The findings of the current study indicate a strong agreement between mRNA expression as quantified by RT-qPCR and HER2 IHC scores, although there was still a substantial proportion of discordant HER2 results between both methods. Future large-scale studies comprising clinical outcome are needed to determine the added (or even superior) value of mRNA quantification in nonamplified HER2 BC.

The authors thank Mieke R. van Bockstal, MD, PhD; Glenn Broeckx, MD, PhD; Cecile Colpaert, MD, PhD; Shusma C. Doebar, MD, PhD; Marieke C. H. Hogenes, MD, PhD; Esther Koop, MD, PhD; Kathleen Lambein, MD, PhD; Dieter J. E. Peeters, MD, PhD; Renata H. J. A. Sinke, MD, PhD; Johannes Bastiaan van Brakel, MD, PhD; Jose van der Starre-Gaal, MD, PhD; Bert van der Vegt, MD, PhD; Koen van de Vijver, MD, PhD; Celien P. H. Vreuls, MD, PhD; Willem Vreuls, MD, PhD; and Pieter J. Westenend MD, PhD, for their participation in the scoring of the slides; and Hein Sleddens, MSc, and Ludo Uytdewilligen, MSc, for performing and scoring the fluorescence in situ hybridization assays of our previous interobserver study.