Immunohistochemical Surrogates for Molecular Classification of Breast Carcinoma: A 2015 Update Open Access

Although commonly used, IHC analysis is subjected to preanalytic, analytic, and postanalytic variables, especially in tissue handling and fixation, antibody sources and clones, and evaluation by pathologists. In recent years, standardization in IHC analysis for breast cancer has improved significantly, especially with the publication of the guidelines for ER, PR, and HER2 testing.^50,51  Immunohistochemistry procedures and scoring methods for each biomarker are briefly described here.

ER is a nuclear protein with 1 DNA-binding domain and 2 activation function domains. A total of 65% to 75% of IBCs express ER, with ER being a major contributor to its development. Clinically, ER-expressing IBCs are usually better differentiated, have a favorable prognosis, and respond to hormonal therapy. The American Society of Clinical Oncology/College of American Pathologists (ASCO/CAP) guideline recommends that ER be considered positive if 1% or more of tumor cells have nuclear staining of any intensity.⁵⁰  Allred scoring and H-scoring are two other commonly used systems for ER and PR evaluations.^52,53 

Similar to ER, PR is also a transcription factor, largely controlled by ER and to some degree by growth factors as well, and is expressed in 55% to 65% of IBC. In most cases, PR coexpresses with ER. Clinically, loss of PR suggests a more aggressive behavior and less response to hormonal therapy. The ASCO/CAP guideline recommends that PR be considered positive if 1% or more of tumor cells have nuclear staining of any intensity.⁵⁰ 

HER2 encodes a transmembrane tyrosine kinase receptor that binds to its extracellular signals and initiates a signaling cascade mediating cell proliferation and survival. About 12% to 20% of IBCs either overexpress the HER2 protein and/or have HER2 gene amplification, resulting in aggressive tumor growth and poor clinical outcome. Its expression is also indicative of a potential response to anti-HER2 therapy. The ASCO/CAP guideline recommends that HER2 be defined as positive if 10% or more of tumor cells exhibit strong uniform membrane staining.⁵¹ 

Ki-67 is a nuclear marker for proliferation expressed in all phases of the cell cycle except G₀. Clinically, higher Ki-67 expression is usually associated with a higher grade and more aggressive behavior. Nuclear staining of any intensity is evaluated. Currently, there is no standardized cutoff value for Ki-67.^54–56  A meta-analysis of 46 studies with 12 155 patients has shown that high Ki-67 labeling correlated with increased relapse and decreased survival. However, the threshold designated as high Ki-67 labeling ranged from 3.5% to 35%.⁵⁷  A 13.25% of Ki-67 labeling was proposed as the cutoff to separate luminal B tumors from luminal A tumors based on PAM50-defined subtyping in 357 patients, even though the misclassified rate was as high as 25%.²⁵  Although a 14% cutoff was endorsed in St Gallen in 2011,⁵⁸  most of the panel voted for a cutoff of greater than 20% as the new cutoff in the 2013 St Gallen conference.³⁰  Also, because there is great variability within any given tumor for Ki-67 labeling, the actual score largely depends on the areas sampled and analyzed.

CK5 is a high–molecular weight cytokeratin and is expressed in normal myoepithelial cells. Its expression in breast cancer is associated with expression of EGFR, Ki-67, p53, and increased cytogenetic abnormalities⁵⁹ ; CK5/6 is often expressed in BRCA1-related breast cancers.⁶⁰  Nielsen et al²⁸  report that a panel of 4 markers (ER, HER2, CK5/6, and EGFR) accurately identifies BLBC. We have demonstrated that CK5/6 gave a higher detection rate than CK14 for BLBC.⁶¹  The cutoffs for its positivity in the literature range from any positive cytoplasmic staining to 20% of tumor cells.⁶² 

EGFR, also known as HER1, is a member of the HER family. Although the IHC expression of EGFR is not indicative of eligibility for current EGFR-targeted therapy, it has enhanced the identification of BLBC significantly because of its higher expression and easier scoring than that of CK5.²⁸ 

AR is a steroid nuclear receptor and is expressed more frequently than ER in breast cancer.⁶³  Although the precise role of AR in breast carcinogenesis is not clear, the possibility of it serving as a therapeutic target, especially in TNBC, is of significant interest.³⁵  Clinically, AR positivity is associated with better prognosis and, more recently,²¹  associated with a lower recurrence score using the 21-gene test.⁶⁴ 

By controlling the cell cycle and inducing apoptosis when cell damage is beyond repair, p53 is a vital regulator of genomic stability. In normal cells, p53 protein has a very short half-life, with degradation in minutes. However, a missense mutation of p53 results in protein stabilization and accumulation in the nucleus.⁶⁵  Expression of p53 has been shown to be associated with poor outcome for patients with IBC in some studies, but has not been shown in others,^66–68  particularly when it is in combination with other markers, such as bcl-2⁶⁹ or HER2.⁷⁰  The different antibodies used in studies seem to produce different results as well.⁷¹  Currently, most investigators use 10% of nuclear staining in tumor cells as the cutoff for p53 positivity.

Most ER-positive tumors fall into the luminal A and luminal B subtypes; however, they do not capture all of the biologic complexity of these tumors.^72,73  One recent study suggests that there are 6 distinct subgroups of ER-positive IBC, indicating the highly heterogeneous nature of these tumors.⁷⁴ 

Luminal A breast cancers account for about 30% to 40% of all IBC. They demonstrate the highest level of expression for ERα gene, GATA-binding protein 3 (GATA3), X-box–binding protein 1, trefoil factor 3, hepatocyte nuclear factor 3, and estrogen-regulated LIV-1.^13–15  They do not overexpress HER2, and about 13% of luminal A tumors have a p53 mutation and 45% show a PIK3CA mutation.^13–15,75  The GATA3 mutation mainly occurs in this subtype and is associated with better prognosis.⁷⁶  Morphologically, most luminal A tumors are well differentiated carcinomas of no special type, tubular carcinomas, classical lobular carcinomas, mucinous carcinomas, and neuroendocrine carcinomas.⁷⁷  The 2013 St Gallen guidelines propose that a low recurrence score from the Oncotype DX test also suggests a luminal A subtype.³⁰ 

Luminal A tumors often have an IHC profile of high ER and PR expression, negative HER2, and low Ki-67. The original IHC surrogates were ER and/or PR positive, and HER2 negative.^11,78  Subsequent studies showed that a Ki-67 of 14% was the cutoff point to separate luminal A from luminal B subtypes.²⁵  More recently, a 20% cutoff point of PR to separate luminal A and B subtypes has been proposed,²⁶  and this was substantiated in a separate study.⁷⁹  Based on currently available data, the most commonly used IHC surrogates for luminal A subtype are ER⁺, PR of 20% or greater, HER2⁻, and Ki-67 of less than 14%.

Luminal B subtype accounts for about 20% to 30% of all IBC. It demonstrates low to moderate expression of the luminal-specific genes, including the ER cluster, but not HER2. Compared with luminal A subtype, it shows a higher frequency of p53 mutations (32%) and a lower frequency of PIK3CA mutations (32%).^13–15,75  Morphologically, this group of tumors is less well differentiated and consists mostly of invasive ductal carcinomas of no special type, and also some invasive micropapillary carcinomas.⁷⁷ 

By IHC analysis, these tumors show a lower level of ER expression, a lower level of or negative PR expression, and a higher level of Ki-67 labeling. However, molecular studies have shown that luminal B is not simply a more proliferative variant of luminal A, because both luminal A and B cancers have their own specific oncogenic drivers.⁸⁰  The IHC definition for this subgroup is less well defined. Originally, ER⁺ and PR⁻,⁸¹  or ER⁺ and HER2⁺,⁸²  was used, but later most investigators used ER⁺ and HER2⁻ to describe this subtype if the score of Ki-67 was 14% or higher,^25,52  or higher than 20%.³⁰  More recently, a cutoff of less than 20% for PR²⁶  has been proposed. Cancello et al⁸³  reported that loss of PR increases the risk of relapse for patients with luminal B tumors. Prior studies also showed that the loss of PR is associated with poor prognosis, decreased responsiveness to endocrine therapy, a higher recurrence score with the Oncotype DX test, and a greater chance of bone metastasis.^{6–8,84–86}  Indeed, the latest St Gallen guideline proposed that a high Oncotype DX recurrence score is suggestive of luminal B–like subtype.³⁰  More recent studies using IHC4 assay (ER, PR, HER2, and Ki-67) and Magee Equations have also emphasized the important role of PR levels in ER-positive breast cancer.^87,88 

Most recently, data from a study with more than 9000 patients with an 8-year follow up indicated that combining Ki-67 and PR levels could better separate luminal B subtype from luminal A subtype.²⁸  The authors defined luminal A as ER⁺, HER2⁻, and Ki-67 of less than 14% with any PR, or Ki-67 of 14% to 19% with PR greater than 20%; and luminal B as ER⁺, HER2⁻, and Ki-67 of greater than 14% with any PR, or Ki-67 of 14% to 19% and PR less than 20%. This proposal was supported by another large study of ER⁺, HER2⁻ breast cancers, in which PR negativity and high Ki-67 level predicted early relapse.⁸⁹  Based on currently available data, the most commonly used IHC surrogates for luminal B subtype are ER⁺, HER2⁻, and PR less than 20% or Ki-67 14% or greater.

This subgroup consists of 12% to 20% of all IBC. HER2-positive tumors are very heterogeneous at the molecular level.^9,90  One study showed that HER2/HER3 heterodimers and p21 expression predicted response to adjuvant trastuzumab in HER2-positive tumors.⁹¹  Among HER2-positive tumors, hormone receptor status is the most important factor affecting patients' survival and treatment response. A recent study from Fountzilas et al⁹²  clearly demonstrated that luminal HER2 (ER⁺, PR⁺, HER2⁺) and HER2-enriched subtypes (ER⁻, PR⁻, HER2⁺) are clinically distinct, with different survival curves and metastatic patterns. Thus, HER2-positive tumors should be divided into two subtypes: luminal HER2 subtype (ER and/or PR positive/HER2 positive) and HER2-enriched subtype (ER and/or PR negative/HER2 positive).^92–94 

About half of HER2-positive IBCs fall into this luminal HER2 subtype. They are ER positive, although often at lower levels compared with HER2-negative tumors.^95–97  Tumors from this subtype have a higher expression of such genes as GATA3, BCL2, and ESR1, and a higher frequency of GATA3 mutations.^13–15,75  Morphologically, they are mostly grade 2 to 3 tumors (often with low or negative PR), relapse early, show frequent nodal metastasis, and have a diminished response to endocrine therapy.^98,99  One large study showed that tumors from luminal HER2 subtype have more frequent locoregional recurrence compared with hormone receptor–negative, HER2-enriched subtype (9.8% versus 3.8%).⁹³  Preclinical evidence seems to confirm that the crosstalk between HER2 and ER signaling pathways may contribute to the resistance to endocrine therapy.^100,101  Simultaneous inhibition of both HER2 and ER pathways seems to be more effective than ER inhibition.^102,103 

The luminal HER2 subtype can be further divided into two phenotypes based on PR expression: ER⁺/PR⁺/HER2⁺ (triple-positive cancer) and ER⁺/PR⁻/HER2⁺, each having distinct clinical properties.¹⁰⁴  PR expression correlated inversely with HER2 overexpression, suggesting the loss of PR in ER-positive tumors may be a marker for an activated EGFR/HER2 pathway.^6,105  Patients with ER⁺/PR^−/HER2⁺ tumors have a reduced breast cancer–specific survival compared with patients with ER⁺/PR⁺/HER2⁺.⁸³  A recent study showed that the survival of patients with ER⁺/PR⁺/HER2⁺ tumors was superior to those with ER⁺/PR⁻/HER2⁺ tumors across all stages, confirming that loss of PR is an unfavorable event.¹⁰⁶  These results highlighted the clinical importance of separating ER⁺/PR⁺/HER2⁺ tumors from ER⁺/PR⁻/HER2⁺ tumors. Based on currently available data, the most commonly used IHC surrogates for luminal HER2 subtype are ER and/or PR⁺/HER2⁺, which may be further divided into ER⁺/PR⁺/HER2⁺ and ER⁺/PR⁻/HER2⁺ subtypes.

By GEP, this subtype is characterized by low to absent gene expression of ER and several additional transcriptional factors expressed in the luminal/ER⁺ cluster. It has high expression of several genes in the ERBB2 amplicon at 17q22.24, including ERBB2 and GRB7. About 71% of the tumors in this subtype have a p53 mutation and activation of receptor tyrosine kinase pathways, such as FGFR4, EGFR, and HER2; and 39% of tumors have a PIK3CA mutation.^13–15,75  There is an incomplete overlap between the HER2 groups as defined by molecular (GEP-based) and clinical (IHC-based or fluorescence in situ hybridization–based) criteria.¹⁰⁷  One study with GEP for clinically HER2-positive tumors identified 3 distinct subtypes with mixed stage, histologic grade, and ER status.¹⁰⁸  Morphologically, besides invasive ductal carcinoma of no special type, special types of ER-negative tumors, including some of the apocrine carcinomas and pleomorphic invasive lobular carcinomas, also belong to this subtype. Based on currently available data, the most commonly used IHC surrogates for HER2-enriched subtype are ER⁻, PR⁻, HER2⁺.

Triple-negative breast cancer is defined by the lack of expression for ER, PR, and HER2 by IHC analysis,¹⁰⁹  and has significant overlap with BLBC as defined by GEP. However, the concordance is not perfect. TNBC patients are usually younger,¹¹⁰  with higher-grade tumors¹¹¹  and a higher risk of distant recurrence and death within the first 3 to 5 years after diagnosis.¹¹²  Most importantly, there is no targeted therapy available for TNBC at the present time. Morphologically, most TNBCs are ductal carcinoma of no special type; others include special types of ER-negative tumors, such as adenoid cystic carcinoma, secretory carcinoma, metaplastic carcinoma, and carcinoma with medullary features, each with a distinctive morphology and clinical behavior.

About 15% of all IBCs belong to this subtype, and there is a great diversity in the histologic features, mutation profiles, response to chemotherapy, metastatic behavior, and survival.¹¹³  Basal-like breast cancer, defined by GEP, shows high expression of many of the genes characteristic of breast basal epithelial cells, such as keratins 5 and 17, laminin, and fatty acid–binding protein 7. Basal-like breast cancer can also be defined using IHC surrogates, including ER, PR, HER2, CK5, and EGFR. There have been p53 mutations reported in 82% of these BLBCs, and most are positive for keratins 5/6 and 17. Morphologically, BLBCs are frequently high grade and large size, with pushing borders, with a poor Nottingham Prognostic Index, and a high rate for local recurrence and distant visceral organ metastasis, especially within the first 5 years. Notably, there is a small group of BLBCs that are low grade and of good prognosis, including adenoid cystic carcinoma and secretory carcinoma with characteristic molecular changes. Secretory carcinoma has the t(12;15)(p13; q25) ETV6-NTRK3 fusion gene, and adenoid cystic carcinoma has the t(6;9)(q22-23; p23-24) MYB-NFIB fusion gene.

Further study on TNBC has confirmed the heterogeneous nature of this group. Lehmann et al¹²  further divided TNBC cancer into 6 different subgroups based on a detailed GEP study, namely: basal-like 1 and 2 subtypes; immunomodulatory subtype; mesenchymal and mesenchymal stem–like subtypes; and luminal androgen receptor subtype. Each subtype has its specific gene expression pattern and response to its specific targeted therapy.¹²  Responses to neoadjuvant chemotherapy are also different, with basal-like 1 having the highest complete pathologic response rate (52%) and basal-like 2 and luminal androgen receptor tumors having the lowest complete pathologic response rates (0% and 10%, respectively).¹¹⁴ 

Triple negativity is the most commonly used IHC surrogate for basal-like tumors.¹¹  Immunohistochemical analysis of BLBC cases that had been defined by GEP demonstrated that added IHC testing for CK5/6 and EGFR was better able to define this subtype than triple negativity alone.^28,115  Based on currently available data, the most commonly used IHC surrogates for BLBC are ER⁻, PR⁻, HER2⁻, CK5⁺, and/or EFGR⁺.

There is a small group of TNBCs that are also negative for CK5 and EGFR.¹¹⁶  Rakha et al¹¹⁷  showed that this subgroup of tumors is less likely to be associated with a BRCA1 mutation, and they have better breast cancer–specific survival and disease-free survival compared with BLBC.

Molecular apocrine breast cancers are ER-negative, AR-positive tumors with apocrine histology; they can be either TN or HER2 positive^21,22  and are associated with early recurrence.²³  The commonly used IHC surrogates are ER⁻, PR⁻, and AR⁺.³⁶  Recently it was shown that although all MABCs are ER⁻, AR⁺, and FOXA⁺ by GEP, only 93%, 58%, and 90% of these tumors are ER⁻, AR⁺ and FOXA⁺, respectively, by IHC analysis.¹¹⁸  Because 67% and 57% of MABCs are positive for HER2 and GCDFP15, respectively, 94% of all MABCs can be identified by these two markers in ER-negative tumors. Thus, a modified IHC surrogate panel may be ER⁻, PR⁻, AR⁺ plus either HER2⁺ and/or GCDFP15⁺. Clearly more studies need to be done to further characterize this subtype of tumors.

Claudin-low BC has low-to-absent expression of luminal markers and enrichment of epithelial-to-mesenchymal transition markers.^19,20  This subtype is usually TN with poor prognosis and enriched for metaplastic carcinoma and carcinoma with medullary features. An IHC-based study showed that CLBCs (defined by low expression of claudins 1, 3, 4, 7, and 8) are found in 77% of BLBCs, 20% of HER2-positive cancers, and 3% of luminal cancers.¹¹⁶  Also, in a large study using GEP, 36% of CLBCs were ER positive.¹¹⁹  These studies suggest that this subtype consists of a group of heterogeneous tumors.

The normal breast-like subtype of breast cancer, as identified in the original GEP intrinsic subtypes, has the same signature as fibroadenomas and normal breast samples.¹³  They express genes characteristic of adipose tissue and lack of expression of ER and HER2. The clinical significance of these tumors remains to be determined because of the lack of studies on this subtype. In fact, in a study using the supervised risk predictor (PAM50) of breast cancer intrinsic subtypes, the normal breast-like subtype was not included.¹²⁰  Also, no single case of normal breast-like subtype was found in a large series of grade 3 breast cancer samples, in which each case had undergone microdissection and was made up of more than 90% of tumor cells. These data suggest that this normal breast-like subtype may be a technical artifact because of the presence of a high volume of normal breast tissue in the original 2000 study.¹²¹  Thus, we decided not to include this subtype in the Table and the Figure.

• 1.

  Immunohistochemistry-based MC is not equivalent to intrinsic subtypes as defined by GEP. Although IHC-based MC could separate breast cancer into subgroups that have similar clinical outcomes compared with those defined by GEP, they are not identical.^107,122  After comparing a large database with more than 1500 cases, Cheang et al¹²²  demonstrated that although there were overlaps between subtypes defined by GEP and IHC analysis, they were not identical. For example, only 73% of IHC-based TNBCs belonged to intrinsic basal-like subtypes, whereas another 17% belonged to intrinsic HER2-enriched subtype. Likewise, only 65% of clinically HER2-positive tumors belonged to intrinsic HER2-enriched subtype, whereas the others belonged to intrinsic luminal B (20%), basal-like (14%), and luminal A (7%) subtypes.¹⁰⁷ 

• 2.

  Issues with cutoff for ER and Ki-67:

• A.

  ER: Several recent studies suggested that IBCs with low ER expression (1%–10%) were more similar molecularly and biologically to BLBC than to luminal cancers,^123,124  whereas other authors reported that only 18% of low–ER expression (1%–9%) tumors belonged to BLBC, with the remaining tumors belonging to intrinsic luminal A or B subtypes (44%).¹²²  Indeed, IBCs with low ER expression were uncommon, accounting for 2% in our cohort¹²⁵  and displaying pathologic features and clinical outcomes intermediate between ER-negative expression (<1%) and higher ER expression (11%–70%) tumors.^125,126  So far, the behavior of the low-ER (1%–10%) IBC is far from clear, with somewhat conflicting results in the literature. Thus, using the current 1% cutoff as endorsed by ASCO/CAP for IHC-based MC appears to be sensible at the present time.

• B.

  Ki-67: The cutoffs for Ki-67 in the literature range from 3.5% to 35%.⁵⁷  A cutoff of 13.25% of Ki-67 labeling was proposed to separate luminal B BC from luminal A BC based on data from intrinsic subtyping.²⁵  The St Gallen breast expert panel had endorsed the 14% cutoff for Ki-67 in 2011.⁵⁸  However, most of the panel voted for a cutoff of greater than 20% as the new cutoff in the 2013 conference.³⁰  One study proposed the IHC surrogates for luminal A tumors as ER⁺, HER2⁻, and Ki-67 less than 14% with any PR, or Ki-67 14% to 19% with PR greater than 20%; and luminal B tumors as ER^+, HER2⁻, and Ki-67 greater than 14% with any PR, or Ki-67 14% to 19% and PR less than 20%.²⁸ 

• 3.

  Awareness of intratumor heterogeneity in breast cancer: Although in most situations, one can assign an IBC into one of the subgroups, there is still considerable intratumoral heterogeneity.^127,128  One recent preclinical study showed that ER-positive tumors admixed with a hormone receptor–negative, CK5-positive basal-like subpopulation will respond better to a combination of endocrine and EGFR inhibitors.¹²⁹  Understanding the intratumor heterogeneity of BC will be critical in understanding the MC of IBC.

Clearly, MC of IBC has enhanced our understanding of breast carcinogenesis significantly in recent years. The capacity to define subtypes of breast cancer provides a framework for understanding the mechanism of breast carcinogenesis, and opportunities for improving therapeutic intervention. The IHC surrogates have been shown to be useful for advancing the understanding of the prognostic and predictive values of MC. The most current understanding of MC is summarized in the Table and illustrated in the Figure. Much more work needs to be done, especially in the standardization of IHC analysis and scoring for each biomarker, standardization of the definition for each classification, and the continued addition of newly discovered biomarkers, before MC can be used routinely in the clinical setting.

We wish to thank Ms. Mary Jackson for her excellent assistance in editing this paper.