Triple Negative Breast Cancer Subclassified by Immunohistochemistry: Correlation with Clinical and Pathological Outcomes in Patients Receiving Neoadjuvant Chemotherapy

1. Introduction

According to Cancer Research UK, triple-negative breast cancer (TNBC) accounts for approximately 15% of malignancies affecting the breast. Characterized by the absence of estrogen, progesterone and Her2 receptor positivity, it is usually associated with a poor prognosis, and chemotherapy remains the main basis of systemic treatment.[1] In the last decade, the relevant benefits of chemotherapy combined with immunotherapy and targeted therapy for patients with pathogenic germline BRCA mutations have positively impacted outcomes in both curative and palliative settings.[2]

However, TNBC represents a heterogeneous group of several subtypes with distinct drivers, survival outcomes and responses to systemic therapy.[3] Several classifications have been proposed and validated, largely based on genomic evaluation of the tumors.[4] In 2011, Lehmann et al characterized 7 subtypes by gene expression, including one unstable. Basal-like 1 (BL1) is associated with genomic repair pathway deficiencies; basal-like 2 (BL2) is enriched in growth factor signaling-related genes; immunomodulatory (IM) is enriched in immunological signaling; and mesenchymal (M) and mesenchymal stem–like (MSL) are enriched in components and pathways related to cell motility and the luminal androgen receptor (LAR), and androgen and metabolism genes are enriched.[5] Later, in 2015, Burstein described four subtypes, namely, luminal androgen receptor (LAR), mesenchymal (MES), basal-like immune-suppressed (BLIS) and immune-activated (BLIA) subtypes.[6] One year later, the FUSCC classification introduced the immune-modulatory (IM) subtype, along with the known MES, BLIS and LAR subtypes.[7]

All the above classifications are based on molecular profiling, which could be valuable at the patient level for precision medicine approaches once some mutations predict the benefit of targeted therapy. However, for most patients, the treatment options will remain limited. Moreover, technical challenges, such as the need for fresh tissue, the length of time needed to perform the analysis, and the significantly high costs, impact its wide use in standard practice.[8,9]

Immunohistochemistry (IHC) is an established laboratory method for evaluating the presence of antigens that is widely used in breast cancer and is pivotal for tailoring systemic treatments (e.g., the human epidermal growth factor receptor 2 - HER2).[10] Moreover, IHC is significantly less expensive and laborious than gene expression and has been evaluated in several studies as a surrogate marker of established intrinsic genomic-based subtyping assays, as summarized in Table 1. [11,12,13,14,15,16,17,18]Although some studies have provided survival data amongst the subtypes, the effect of chemotherapy on the pathological complete response, if it also serves as surrogate biomarker of survival has not been reported.

In this study, we report the survival and pCR rates of patients who underwent IHC subtyping using our panel and who received neoadjuvant chemotherapy.

2. Results

One hundred eighty-seven patients were included in this analysis; the mean age was 52.43 (SD = 12.79) years, the mean tumor size was 47.64 (SD = 53.61) millimeters, and other demographic characteristics are summarized in Table 4.

For immunohistochemistry, staining for all 4 markers was possible in 98.4% (184) of the patients and partially in 3 patients due to tissue exhaustion. We successfully identified the characteristics of 4 subtypes in 96.3% (183) of patients. Overall, the frequency of the IHC-basal subtype was 61% (114). The percentages of basal subtypes identified by IHC-BLIS and IHC-BLIA were 31% (57) and 10.9% (20), respectively. The second most common subtype was IHC-MES 12.5% (23), followed by IHC-LAR 12.0% (22) and IHC-mixed 11.4% (21). (Table 2)

Among the 28 (15%) patients who achieved pCR, 75% exhibited an IHC-basal feature, 14.28% exhibited an IHC-mesenchymal feature, and 7.4% exhibited an IHC-LAR feature. The pCR rate in the IHC-BLIA group was the highest (30%), and that in the IHC-LAR group was the lowest (4.5%). (Table 3)

Patients who achieved pCR had a significantly better median disease-free (mDFS) survival NR (not reached) (CI: >50%) vs. no-PCR 30 m (months) (CI: 18-NR) (p=0.00022). Overall, it was also significantly greater in patients who achieved pCR according to the NR (CI: >50%) than in those who did not achieve pCR according to the 58-m (CI: 40-NR) no-PCR (p=0.00018). (Supplementary Figures S1 and S2).

Most of the recurrences occurred in the first 20 months following surgery. mDFS did not significantly differ among subtypes (p=0.52). (Figure 1A) IHC-BLIA demonstrated a greater mDFS for patients with NR (CI: 34-NR) than for patients with all other subtypes (CI: 23-NR) (p=0.14). The meDFS was also NR for the IHC-LAR, IHC-mixed and IHC-basal-like groups. ICH-BLIS patients had the second worst mDFS, 50 m (CI: 11-NR), and IHC-MES patients had a markedly short mDFS, 16 m (CI: 6-NR). Interestingly, the mDFS curves for patients who did not achieve pCR were less distinct among subtypes, suggesting that pCR has a crucial role regardless of intrinsic subtype. IHC-MES followed by IHC-Basal-like showed worse survival, respectively 17 m (CI: 6-NR) and 15m (CI: 6-NR). (Figure 1B).

The pattern of recurrence was graphically distinct among subtypes. IHC-BLIS led recurrence in each category, and interestingly, no local or cerebral relapse occurred in patients with the IHC-LAR subtype. (Figure 2)

Overall survival analysis also revealed no significant differences among subtypes (p=0.61).(Figure 1C) IHC-BLIA had a greater mOS, NR (95% CI: NR>50%), than did the other subtypes at 60 m (95% CI: 48-NR) (p=0.081). IHC-BLIA exhibit the graphically highest mOS, with an NR (95% CI: 39-NR) mOS for patients who did not achieve pCR and IHC-MES was associated with a worse survival with 37m (95% CI: 27-NR). (Figure 1D)

3. Discussion

This study is pioneering in demonstrating pCR rates among TBNC subtypes using IHC. We found numerically higher pCR rates in the IHC-BLIA (30%) subgroup and a trend toward longer DFS and OS, while in the IHC-MES subgroup, survival was worse despite the pCR rate (17.4%) being closer to that of the whole population analyzed (15.0%).

The basal-like subtype was predominant in our sample, with 61.5% of the tumors harboring a basal component. In 2010, Oakman C et al reported higher rates of the BL phenotype in TNBC patients (71-91%).[19] Interestingly, Lehmann et all in 2011 reported that 49% of BL cases were characterized by intrinsic subtyping, but IHC seems to indicate a higher percentage (88%), and very likely, only half of these cases would correlate with the intrinsic BL subtype.[5] We observed rates somewhat closer to those reported by Zhao et al. and Jing Liang et al.[14,16] Identifying basal-like TNBC is clinically relevant, as data showed a benefit from adjuvant chemotherapy for patients with non-basal-like tumors and residual disease following surgery, despite the use of neoadjuvant chemotherapy. [20,21]

We identified a predominance of IHC-BLIS (31.0%) over IHC-BLIA (10.9%) in our population, which likely contributed to the overall modest pathological response rate of (15.0%). Although our IHC-BLIS rates are somewhat comparable to the values found by the authors performing IHC subtyping, the overall percentage of patients with the IHC-BLIA subtype is markedly lower.(Table 1) On the other hand, we found a markedly higher pCR rate in IHC-BLIA (30.0%) than the pCR rate in IHC-BLIS (14.0%). Based on recent publications, the new standard of care for early and high-risk TNBC patients includes immunotherapy alongside a taxane-platinum-based dose-dense chemotherapy, regardless of the presence of immune receptors/markers.[22] However, data are still needed to understand the efficacy of these regimens in TNBC patient subpopulations.

Immuno-inflamed occurred solely at the IHC-IM subtype according to Jing Lian et al. (2022), but a significant proportion of the other subtypes are immune-excluded, which raises the question of whether those subjects would benefit from immunotherapy in clinic.[16]

IHC-MES was the second most common subtype, and numerically, it was associated with worse mDFS and mOS, which is consistent with the findings in the literature. [13,14,15] Unfortunately, the benefit of conventional adjuvant systemic treatment for this subtype is often poor, if any, given its specific molecular characteristics, which has motivated researchers to consider regimens mimicking sarcomatous disease. [23] Moreover, given the preclinical and clinical activity of mTOR, PI3K, SRC/ABL and angiogenesis inhibitors, we believe that tailored treatment for this subtype could be warranted.[2]

The IHC-LAR subtype had the lowest pCR rate (4.5%) and was the third most common subtype in our sample. A long-standing effort has been made to offer anti-androgen targeting for this subtype, with no translation in registrational approvals. The main challenges might include the unreliable behavior of AR as a driver in TNBC, compared to that of ER in HR+ disease, as well as the cutoff of AR-receptor intensity to trigger intervention and the high-bar endpoints chosen in clinical trials to evaluate the AR-targeting effect.[24] As mentioned above, the current best adjuvant treatment for patients with residual disease is capecitabine; however, for specific subtypes, such as LAR, the benefit is unclear. Therefore, to answer this question and by leveraging our knowledge of prostate cancer, where androgen deprivation therapy plus androgen receptor pathway inhibitors can be combined with docetaxel, we believe that a cohort of patients with IHC-LAR in a potential umbrella trial for intrinsic subtypes of TNBC should considered to receive adjuvant capecitabine plus ARPI.[25]

The pattern of recurrence we observed in our IHC-based subtyping cohort aligned with that expected in the literature.(Table 1) Notably, brain relapse, a known independent factor of poor prognosis, occurred in 8% of the patients with IHC-MES, 14% in IHC-BLIS, and approximately 10% of those with the other IHC subtypes, but interestingly no patients with IHC-LAR experienced brain relapse. There is a longstanding discussion on how to optimize relapse monitoring in high-risk breast tumors, but no clear consensus has been reached. Our data could assist in identifying patients with a greater risk of recurrence, which could deserve more individualized monitoring given specific patterns of disease relapse and therefore might benefit from brain and visceral imaging, as well as promising emerging techniques, such as circulating tumor DNA, in addition to the current recommended approaches.[26]

Our study demonstrated a low percentage of unclassifiable 2.2%, and IHC-mixed 11.4% compared to the other studies mentioned in Table 1. In a study conducted by Yoo in 2022, the majority of patients with IHC-unclassifiable had presented with mesenchymal subtype on genomic classification.[15] Interestingly, both Zhao et al., Leeha et al. and Hu et al, did not reported mixed subtypes by using their IHC panels.[14,17,18] We believe IHC-unclassifiable and IHC-mixed should be expected given the IHC assay limitations and support the heterogeneous features of a significant proportion of TNBCs. [27]

Although the results of this study are promising, they should be interpreted in light of these limitations. Our retrospective design imposes challenges in survival analysis. There is no consensus about the use of an IHC subtyping panel, and IHC staining was performed utilizing archival tumor tissue from 2008 onward, which could have impacted the performance of the assay and requires validation using an established method for intrinsic genomic subtyping.

4. Materials and Methods

4.1. Sample Selection and Outcome Definitions

Patients with localized or locally advanced TNBC who were treated with neoadjuvant chemotherapy followed by curative surgery at the same institution between January 2008 and December 2017 and who had available biopsy and surgical specimens met the inclusion criteria for this retrospective cohort.

From medical files, clinical and pathological data were collected to construct a database using an electronic case report form (supplementary file S1).

For this study, TNBC was defined as estrogen and progesterone receptor < 1% and HER2-negative +1 or +2 without fish amplification. Pathological complete response (pCR) was defined as the absence of invasive cancer on surgical specimens from the breast and axilla (ypT0/Tis pN0). Disease-free survival (DFS) was defined as the time between the day of the cycle of one day of chemotherapy regimen administration until cancer recurrence or cancer-related death. Overall survival (OS) was defined as the time between the day of the cycle one day after chemotherapy regimen administration until death related to any cause. In the case of loss to follow-up, the appointment date at the NCI, which contains information about disease and survival, will be used to censor the patient for the analysis.

2.2. IHC Analysis and Interpretation

Tissue samples were converted into histological sections on previously salinized slides. Evaluation of the cellular atypia pattern was performed by hematoxylin-eosin (HE) staining. The evaluated tumor area was selected by a pathologist, and the in situ regions within the invasive tumors were also delineated to allow a better evaluation of the stains. Tissue sections were immunoassayed with a Polymer Detection System (RE7150-K, Leica Biosystems Newcastle Ltd., Balliol Business Park West, Benton Lane, Newcastle Upon Tyne NE12 8EW, United Kingdom) according to the protocol established by the manufacturer. Primary antibodies were incubated with the tissues for 18 h at 4˚C at different dilutions determined by titration, as shown in Table 4.

The antibody panel was adapted from studies reported in the Table 1 and interpretation of the IHC results was performed by the author F.R., as shown in Table 3. The scoring was reviewed by the senior authors E.A., and the final scores agreed between F.R. and E.A..

Table 5. Immunohistochemistry analysis and interpretation.

IHC-subtype			Antibody results
			IDO-1	FOX-C1	AR	Claudin-3
Basal-like unspecific (BL)			IDO-1 ≤ 10% and FOX-C1 < 4or IDO > 10% and FOX-C1 ≥ 4		< 10%	Weak (W), M or I
Basal like immune-suppressed (BLIS)			≤ 10%	≥ 4	< 10%
Basal like immune-activated (BLIA)			> 10%	< 4	< 10%
Luminal androgen receptor (LAR)			≤ 10%	< 4	≥ 10%
Mesenchymal (MES)			any	any	any	0
Mixed (criteria for ≥ 2 subtypes)	BLIS	LAR	≤ 10%	≥ 4	≥ 10%	W, M or I
	BLIA	LAR	> 10%	< 4	≥ 10%
	BL	LAR	IDO-1 ≤ 10% and FOX-C1 < 4or IDO > 10% and FOX-C1 ≥ 4		≥ 10%
	MES	LAR	any	any	≥ 10%	0
Unclassifiable			≤ 10%	≤ 2	< 5%	W, M or I

Table 5. Immunohistochemistry analysis and interpretation.

IHC-subtype			Antibody results
			IDO-1	FOX-C1	AR	Claudin-3
Basal-like unspecific (BL)			IDO-1 ≤ 10% and FOX-C1 < 4or IDO > 10% and FOX-C1 ≥ 4		< 10%	Weak (W), M or I
Basal like immune-suppressed (BLIS)			≤ 10%	≥ 4	< 10%
Basal like immune-activated (BLIA)			> 10%	< 4	< 10%
Luminal androgen receptor (LAR)			≤ 10%	< 4	≥ 10%
Mesenchymal (MES)			any	any	any	0
Mixed (criteria for ≥ 2 subtypes)	BLIS	LAR	≤ 10%	≥ 4	≥ 10%	W, M or I
	BLIA	LAR	> 10%	< 4	≥ 10%
	BL	LAR	IDO-1 ≤ 10% and FOX-C1 < 4or IDO > 10% and FOX-C1 ≥ 4		≥ 10%
	MES	LAR	any	any	≥ 10%	0
Unclassifiable			≤ 10%	≤ 2	< 5%	W, M or I

For positive controls, tissues suggested by the antibody manufacturer’s datasheets were used. The reaction was visualized using diaminobenzidine (DAB), followed by hematoxylin counterstaining. Negative controls were prepared without the primary antibody. The positivity of the staining was analyzed in ten random fields and determined by manual counting using ImageJ^® software according to the preestablished equation below.

$$\frac{\sum _{\mathit{m}=1}^{10}\mathit{S}\mathit{c}}{\sum _{\mathit{t}=1}^{10}\mathit{T}\mathit{c}}\times 100=\mathit{p}$$

Sc represents the number of stained cells in each field, where s = [1, 2, ∙∙∙, 10]; Tc represents the total number of cells in each field, where t = [1, 2, ∙∙∙, 10]; and p represents the mean percentage of positivity.

2.3. Statistical Analysis

Continuous variables are shown as the mean and standard deviation, and categorical variables are shown as the total value and percentage of the total value.

The time to overall survival was computed as the time of diagnosis until death or the last patient contact, and the disease-free survival time was computed as the time of surgery until recurrence or last patient contact.

The Kaplan‒Meier method was used to estimate survival curves, and the log rank test was performed to evaluate whether there were differences in covariate levels via univariate analysis. The significance level adopted to reject the null hypothesis was alfa equal to 0.05. A bar plot was used to demonstrate the frequency of subtypes at each site.

5. Conclusions

We successfully demonstrated that an affordable and widely reproducible panel of IHC assays can be used to subtype TNBC. Moreover, we were the first to demonstrate the pCR rate among IHC subtypes.