Immunohistochemical and molecular profiles of heterogeneous components of metaplastic breast cancer: a squamous cell carcinomatous component was distinct from a spindle cell carcinomatous component

Metaplastic breast carcinoma (MBC), a category of breast cancer, includes different histological types, which are occasionally mixed and heterogeneous. Considering the heterogeneity of cancer cells in a tumour mass has become highly significant, not only from a biological aspect but also for clinical management of recurrence. This study aimed to analyse the immunohistochemical and molecular profiles of each MBC component of a tumour mass. Twenty-five MBC tumours were histologically evaluated, and the most frequent MBC component (c) was squamous cell carcinoma (SCC), followed by spindle cell carcinoma (SpCC). A total of 69 components of MBC and non-MBC in formalin-fixed paraffin-embedded sections were examined for 7 markers by immunohistochemistry. SCC(c) were significantly PTEN negative and CK14 positive, and SpCC(c) were significantly E-cadherin negative and vimentin positive. Multivariate analyses revealed that immunohistochemical profiles of normal/intraductal (IC)(c), no special type (NST)(c), and MBC(c) differed; moreover, SCC(c) and SpCC(c) were distinctly grouped. PTEN gene mutation was detected only in SCC(c) (2/7), but not in SpCC(c). Next-generation sequence analyses for 2 cases with tumours containing SCC(c) demonstrated that PTEN gene mutation increased progressively from IC(c) to NST(c) to SCC(c). In conclusion, the immunohistochemical and molecular profiles of the SCC(c) of MBC are distinct from those of the SpCC(c). Supplementary Information The online version contains supplementary material available at 10.1007/s12672-024-00950-0.

Understanding the heterogeneity of cancer cells in a tumour mass has become increasingly important.The morphological heterogeneity observed by histology has attracted the attention of pathologists from a taxonomic point of view, in the past.At the present time, we can understand the process of carcinogenesis via integrated information of phenotypic and genetic knowledge for cancer cells.One such example is mixed ductal and lobular breast cancer.Although ductal carcinoma and lobular carcinoma are classified as different histological types, recent investigations have revealed that intimate mixtures of ductal and lobular carcinoma cells might share genetic alterations and originate from common ancestor cells [13,14].Such biological and pathological understanding of heterogeneity is also critical for adequate treatment and clinical management for recurrent lesions, because a tumour mass containing heterogeneous components is associated with increased risk of remnant and recurrent cancer cells resistant to initial treatments.
MBC tumours comprise rather heterogeneous cancer cells; however, from phenotypic and genetic perspectives, it has not been fully clarified whether cancer cells with a deviated "metaplastic" phenotype arises from carcinoma cells of no special type with acquired additional genetic alterations, or de novo from non-neoplastic epithelial cells.Therefore, we aimed to clarify the lineage among heterogeneous components of an MBC tumour mass by analysing immunohistochemical profiles and gene mutations focusing on each histological component.

Patients
The patients enrolled in this study were selected from the medical records of Nihon University Itabashi Hospital.Inclusion criteria were patients with primary breast cancer diagnosed as MBC who underwent mastectomy in our hospital from 2000 to 2018.Exclusion criteria were receipt of any treatment before surgery.Twenty-five cases were selected, and the clinical data collected were age and presence of recurrence and metastasis; pathological data collected were tumour size, nuclear grade, pathological TNM and stage, vessel invasion and subtype (Supplementary Table 1).This study was conducted in accordance with the Declaration of Helsinki of 1975 and approved by the Institutional Review Board of Nihon University Itabashi Hospital (Approval number: RK-160712-11).This study was a retrospective, observational study, and the consent of patients was waived by the IRB, with an opportunity to decline participation.

Histopathological examination
The resected breast cancer tissues were fixed in 10% neutral buffered formalin and processed appropriately according to routine practice; 4 µm formalin-fixed paraffin-embedded (FFPE) sections were stained with haematoxylin and eosin (HE), and examined by 2 certified pathologists and additional certified pathologist.Histopathological diagnosis was performed according to World Health Organization (WHO) Classification of Tumours of the Breast [15].Based on medical records, no patients had pathological diagnoses of low-grade adenosquamous carcinoma or fibromatosis-like metaplastic carcinoma.Five cases of SpCC, 2 cases of SCC, 4 cases of metaplastic carcinoma with heterologous mesenchymal differentiation and 14 cases of mixed metaplastic carcinoma were included in this study.To examine how metaplastic components develop their phenotypic characteristics, we evaluated heterogeneous components individually, e.g., a no special type (NST) component (c) in mixed metaplastic carcinoma, an intraductal component (IC)(c), and a normal breast tissue (normal)(c), in addition to MBC(c)(SCC(c), SpCC(c), OSS/CAR(c), MPC(c), and PLM(c)) (Fig. 1a-h) (Supplementary  2).All immunohistochemical evaluations and next-generation sequencing (NGS) analyses were performed individually for these histological components.
Immunohistochemical findings were evaluated by certified breast pathologists and classified as positive when > 5% of the tumour cells showed staining.PTEN, LKB1, TGFBR1, vimentin, and CK14 were detected in the cytoplasm; CD44 and E-cadherin were detected at the cell membrane; and p40 was detected in the nucleus (Fig. 1i-p).IHC for p40 was used only to confirm the squamous phenotype indicated by HE staining.
The positive controls used were normal breast tissue for PTEN; colon cancer tissue for LKB1 and TGFBR1; skin tissue for E-cadherin and CK14; adipose tissue for vimentin; and cutaneous squamous cell carcinoma for CD44 and p40.A negative control was performed by omitting the primary antibodies and using control antibodies (mouse IgG and rabbit IgG, Vector Laboratories, Inc., CA, U.S.A.)

PTEN gene mutation analysis
PTEN gene mutations were examined using whole slides of MBC tissues.DNA was extracted from the tissues using the QIAamp DNA FFPE Tissue kit (QIAGEN, Hilden, Germany) according to the manufacturer's instructions.The DNA concentration was assessed using a NanoDrop One instrument (Thermo Fisher Scientific, Waltham, MA, USA).Twelve samples with sufficient DNA (> 10 ng/μl) for analysis were selected for further examination.

Next-generation sequencing analysis
We analysed genetic changes in MBC, including SCC(c), of 2 cases (case no.16, NST(c) and SCC(c); case no. 25, IC(c), NST(c) and SCC(c)) using NGS.Eight-micron-thick FFPE sections were cut and mounted onto glass slides coated with membrane film for laser microdissection.After deparaffinization, the sections were stained with toluidine blue and dried at RT.Each target component was dissected and collected using the PALM® MicroBeam III-N system (Carl Zeiss Microscopy, Munich, Germany).Genomic DNA was extracted using the QIAamp DNA FFPE Tissue kit (QIAGEN) according to the manufacturer's instructions.Initial quality control checks of the FFPE DNA were performed using 1% agarose gel electrophoresis and Qubit dsDNA HS Assay Kit with a Qubit 2.0 fluorometer (Thermo Fisher Scientific).Amplicon libraries were prepared using Ion AmpliSeq™ Library Kit 2.0 (Thermo Fisher Scientific) with Ion AmpliSeq™ Comprehensive Cancer Panel (Thermo Fisher Scientific), which covers all exons of 409 cancer and cancer-related genes, according to the manufacturer's instructions.Library quality was assessed using a 2100 Bioanalyzer (Agilent Technologies, Santa Clara, CA, USA); the libraries were quantified using Ion Library TaqMan® Quantitation Kit and a LightCycler® 480 Instrument II (Roche, Rotkreuz, Switzerland).Sequencing was performed with Ion Proton™ (Thermo Fisher Scientific) using PI Chip (Thermo Fisher Scientific).The sequencing data were analysed with Torrent Suite Software (Ver5.0.4;Thermo Fisher Scientific).Variants were filtered according to the following criteria: total allele coverage greater than 100 and variant allele frequency (VAF) greater than 4%.Germline mutations confirmed in the ClinVar database (https:// www.ncbi.nlm.nih.gov/ clinv ar/) and single-nucleotide variants confirmed in dbSNP (https:// www.ncbi.nlm.nih.gov/ snp/) were excluded

Statistical analyses
Data from IHC results were considered nominal variables.Correlations between IHC results and each histological component were analysed using the chi-square test and a statistical multivariate analysis method for nominal variables, a type III quantification method.P values of less than 0.05 were considered statistically significant.Correlation between IHC markers was analysed by Spearman correlation coefficient analysis.Statistical analyses were performed using BellCurve for Excel® (Social Survey Research Information Co., Ltd., Tokyo, Japan) and R version 4.2.0 (2022-04-22 ucrt), "Vigorous Calisthenics" Copyright (C) 2022 The R Foundation for Statistical Computing Platform: ×86_64-w64-mingw32/ × 64 (64-bit).

Histopathological evaluation of components of MBC tumour masses
Clinicopathological data collected from the medical records for the twenty-five patients with MBC enrolled in this study are summarized (Supplementary Table 1).Results of pathological evaluation for MBC and/or non-MBC components consisting of each MBC tumour are shown in Supplementary Table 2.The most frequently detected MBC(c) was SCC(c) (n = 13), followed by SpCC(c) (n = 8), OSS/CAR(c) (n = 4), MPC(c) (n = 3), and PLM(c) (n = 4); the combination of these components differed among tumours (Supplementary Table 2).As non-MBC(c), normal(c) (n = 15), IC(c) (n = 9), and NST(c) (n = 13) were detected.A total of 69 components, including 32 MBC(c) and 37 non-MBC(c), were included for further examination in this study.

Univariate analysis for the immunohistochemical characteristics of every histological component of MBC tumours
To examine how MBC develops heterogeneous phenotypic characteristics, we evaluated the immunohistochemical features of all 69 components of 25 tumours individually.The percentage of the positive number/total number of components for each SCC(c), SpCC(c), OSS/CAR(c), MPC(c), and PLM(c), NST(c), IC(c) and normal (c) were calculated for seven markers (Fig. 2).Compared to normal(c) (designated as (1) in Fig. 2), the positive percentage for MBC(c) was a significant decrease for PTEN, LKB1, E-cadherin, and TGFBR and a significant increase for vimentin, CD44, CK14, and TGFBR.Compared to IC(c) (designated as (2) in Fig. 2), both NST(c) and MBC(c) showed decreased positivity for PTEN, E-cadherin, and TGFBR and increased positivity for vimentin and CK14.Focusing on each individual MBC(c) (designated as (3) in Fig. 2), we found that a significantly lower percentage of SCC(c) was positive for PTEN (4/13, P < 0.05; red bars in Fig. 2; Supplementary Table 5), with a higher percentage for CK14 (11/13, P < 0.01), compared to other MBC(c)s.A significantly lower percentage of SpCC(c) was positive for E-cadherin (1/8, P < 0.05; green bars in Fig. 2) and a higher percentage was positive for vimentin (7/8, p < 0.05) compared to other MBC(c)s.In addition, CAR/OSS(c) showed a higher positive percentage for vimentin (purple bars in Fig. 2) (4/4, p < 0.05).Thus, SCC(c) and SpCC(c) showed unique immunohistochemical findings compared to other MBC components; SCC(c) was PTEN-negative, CK14-positive (basal-like phenotype) and SpCC(c) was E-cadherin-negative and vimentin-positive (epithelial mesenchymal transition (EMT)-like phenotype).

Multivariate analysis for the immunohistochemical profile of every histological component of MBC tumours
In this study, we used multiple antibodies for multiple components, which may or may not correlate with each other.Therefore, we next evaluated the correlation between antibodies and components by a statistical multivariate analysis method for nominal variables, a type III quantification method.Using this approach, four groups were identified (Fig. 3a): Group (i), including normal(c) and IC(c), was plotted mainly on the minus side of the X-axis; Group (ii), including NST(c) samples, was distributed among all quadrants; Group (iii), including SpCC(c) samples, was plotted mainly in quadrant https://doi.org/10.1007/s12672-024-00950-0I; and Group (iv), including SCC(c) samples, was plotted predominantly in quadrant IV and partially in quadrant I.These results clearly showed that the IHC expression profiles of normal(c) and/or IC(c), NST(c), and MBC(c) were differently distributed, and furthermore, SCC(c) and SpCC(c) showed distinct distributions (Fig. 3a).
We next explored how these immunohistochemical markers are related using Spearman correlation coefficients; PTEN showed positive correlations with LKB1 and TGFBR, and negative correlations with vimentin, CK14, and CD44 (Fig. 3c).E-cadherin correlated negatively with vimentin.

PTEN gene mutations detected in SCC(c)
Our immunohistochemical study demonstrated lower PTEN positivity for both SCC(c) and SpCC(c), though their IHC profiles differed.Therefore, we next sought to examine PTEN gene mutations as a possible cause of decreased PTEN protein expression using 12 available FFPE whole slides of MBC tumours with good preserved DNA quality.Two of 7 tumours containing SCC(c) showed PTEN gene mutations.In one patient (serial case no.14), two missense mutations and two silent mutations were detected; a single deletion was detected in another SCC(c) sample (serial case no.25) (Table 1).These detected PTEN gene mutations are shown in Supplementary Fig. 1.The samples from the remaining 5 patients containing SpCC(c), CAR/OSS(c), MPC(c) or PLM(c) exhibited a wild-type PTEN gene.Thus, PTEN mutations were detected only in the tumour containing SCC(c).

Next-generation sequencing analysis of MBC tumours containing SCC(c)
We next analysed the MBC components that gained gene alterations to evaluate the phenotypic and genetic correlation by NGS.Two candidate MBC tumours were from serial case no.14 and no. 25, both of which were PTEN negative by IHC but positive for PTEN gene alterations (Table 1).However, the DNA quantity of the FFPE samples from case no.14 was not sufficient for NGS analysis.Hence, we used case no.16 in addition to case no. 25 (Table 2 and Supplementary Fig. 2).
For case no.16, eight somatic gene mutations were identified, which were contained in both NST(c) and SCC(c).Specifically, two mutations were identified in KIT, and one mutation was identified in each of the IL7R, NOTCH, PTEN, FLT3, RB1 and TP53 genes (Table 2).The mutations identified in PTEN, RB1 and TP53 are nonsense variants that have been reported as pathogenic in COSMIC database.Within this tissue sample, the pathogenic TP53 mutation was detected at a lower A type III quantification method is one of the standard methods for multivariate analysis of category variables.Quadrant I (right upper) is divided by the X-axis (PTEN negative, CK14 positive, LKB1 negative) and Y-axis (E-cadherin negative, vimentin positive, CD44 negative); quadrant II (left upper) is divided by the X-axis (PTEN positive, vimentin negative, TGFBR positive) and Y-axis (E-cadherin negative, vimentin positive, CD44 negative); quadrant III (left lower) is divided by the X-axis (PTEN positive, vimentin negative, TGFBR positive) and Y-axis (CD44 positive, vimentin negative, TGFBR negative); and quadrant IV (right lower) is divided by the X-axis (PTEN negative, CK14 positive, LKB1 negative) and Y-axis (CD44 positive, vimentin negative, TGFBR negative).Black dots, normal(c); blue dots, intraductal component (IC)(c); yellow dots, invasive ductal carcinoma of no special type (NST)(c); red dots, squamous cell carcinoma (SCC)(c); green dots, spindle cell carcinoma (SpCC)(c); purple dots, carcinoma with cartilaginous/osseous differentiation (CAR/OSS)(c); grey dots, matrix-producing carcinoma (MPC)(c); brown dots, pleomorphic carcinoma (PLM)(c).Group (i), circled by a black line; Group (ii), circled by a yellow line; Group (iii), circled by a green line; Group (iv), circled by a red line.b A panel presenting all of the positive or negative results by immunohistochemistry. Immunohistochemical results for antibodies are shown as red cells for positive, and green cell for negative.Dark grey cells, normal(c); blue cells, intraductal component (IC)(c); yellow cells, invasive ductal carcinoma of no special type (NST)(c); red cells, squamous cell carcinoma (SCC)(c); green cell, spindle cell carcinoma (SpCC)(c); purple cells, carcinoma with cartilaginous/osseous differentiation (CAR/OSS)(c); light grey cells, matrix-producing carcinoma (MPC) (c); brown cells, pleomorphic carcinoma (PLM)(c).c.Correlation between immunohistochemical expression of the markers used in this study analysed by Spearman's rank correlation coefficient.Spearman's correlation coefficients are shown in each corresponding cell.Red cells, positive correlation; blue cells, negative correlation.

Discussion
The present study demonstrated that MBC tumour masses most frequently consisted of SCC(c) and SpCC(c).SCC(c) were significantly PTEN negative and CK14 positive (basal-like phenotype), and SpCC(c) were significantly E-cadherin negative and vimentin positive (EMT-like phenotype).Multivariate analyses showed the immunohistochemical profiles of IC(c), NST(c), and MBC(c) to be distributed differently, and moreover, SCC(c) and SpCC(c) distinctly grouped.PTEN gene mutations were detected only in SCC(c), but not in SpCC(c).Thus, SCC(c) and SpCC(c) presented different immunohistochemical and molecular phenotypes.Based on NGS analyses, PTEN gene mutation was increasingly detected from IC(c) to NST(c) to SCC(c).
There are several lines of evidence showing that PTEN loss contributes to a squamous phenotype.Mouse models addressing the contribution of PTEN loss to SCC in non-mammary organs have been reported.One of such studies for lung SCC showed that biallelic inactivation of Lkb1 and PTEN is required to induce squamous differentiation [16].In addition, Squarize et al. [17] reported that a mouse model of Cowden syndrome, a syndrome associated with germline mutations in PTEN [18], presents a phenotype of acral keratosis that is reversed by induced PTEN expression [17].Another mouse model induced by PTEN deletion shows progression of epidermal hyperplasia, hypergranulosis, hyperkeratosis, mammary adenosquamous carcinoma, adenomyoepithelioma [19,20], and cutaneous SCC, accompanied by hyperactivation of mTOR and production of fibroblast growth factor (FGF) [20].
In terms of studies on human cancer tissue, PTEN gene mutations are more frequently detected in SCC than in adenocarcinoma, e.g., 19% of lung [21], 15% of head and neck [22], 11% of oesophageal [23], and 6%-11% of other cancers [15,24].For breast carcinoma, PTEN loss is observed in 40-50% of cases, and dysfunction of PTEN with frameshift mutations and epigenetic mechanisms is detected in 5-10% [19].Another study of MBC showed that PTEN gene mutations were more frequently detected in SCC (3/9, 33%) than in SpCC (0/10, 0%) and CAR (1/16, 6%) [10].The  One of the unaddressed questions for MBC is whether its heterogeneous phenotypes develop during tumour progression via invasive carcinoma or occur through a de novo mechanism.The interesting findings of this study by NGS analysis were as follows: (1) gene mutations were highly shared between NST(c) and SCC(c) in case no.16 (6 of 8 genes) and between IC(c), NSTC(c) and SCC(c) in case no. 25 (4 of 5 genes); (2) pathologic mutations in PTEN, RB1 and TP53 were increasingly detected with higher VAF in SCC(c) than in NST(c) in case no.16.In case no. 25, no PTEN gene mutation was detected in IC(c) but was increasingly detected in NST(c) and SCC(c).These findings suggest that SCC(c) occurs via NST(c) rather than de novo from non-neoplastic epithelial cells.Considering all the findings together, SCC(c) might develop via NST(c), but not de novo, gaining PTEN gene mutation that affected the SCC(c) phenotype, which process was distinctive from that of SpCC(c).
Possible causes of the decreased PTEN positivity observed in SpCC(c), independent of genetic alterations, are an EMT mechanism, PTEN gene promoter methylation [25], and miRNA involvement [26].PTEN promoter methylation is present in 31.1% of breast carcinoma cases, 64.3% of involve loss of PTEN expression [25].If the pathogenesis of SCC(c) and SpCC(c) is based on different mechanisms, appropriate adjuvant chemotherapeutic strategies would also differ.More effective therapies for each component, such as AKT signaling inhibitors for SCC and EMT inhibitors for SpCC, may be proposed.
A major limitation of this study is its small sample size.The relatively low incidence of MBC makes it difficult to collect sufficient samples with good-quality DNA for molecular analysis over a short duration, because longer duration of formalin fixation increases artefactual modifications in DNA [27], and longer storage of FFPE samples increases fragmentation of DNA strands [28,29].A related issue is that we could not analyse CAR/OSS, MPC or PLM samples in this study cohort.One potential solution for studying tumours with low incidence over a short duration, with good fixation and storage conditions, is to perform a multi-institutional study, and this approach will be considered in the future.

Conclusion
Although SCC(c) and SpCC(c) are occasionally admixed in a single tumour, these components showed distinct immunohistochemical profiles: SCC(c) and SpCC(c) were grouped in the basal phenotype and in the EMT-like phenotype, respectively.Negative expression of PTEN in SCC(c) was accompanied by PTEN gene mutation but not in SpCC(c).NGS analyses demonstrated PTEN gene mutation to be progressively increased from IC(c) to NST(c) to SCC(c).Immunohistochemical and molecular profiles of SCC(c) were distinct from those of SpCC(c) and may originate through NST(c).

Fig. 2 Fig. 3
Fig. 2 The positive percentage for seven markers in every component within MBC samples determined by immunohistochemistry. Comparisons were performed between (1) normal tissue (normal(c)) and carcinoma, (2) intraductal component (IC(c)) and invasive carcinoma, and (3) each metaplastic component (MBC(c)) and other components from all patients.White bars: normal tissue (normal(c)), blue bars: intraductal component (IC(c)), yellow bars: invasive ductal carcinoma of no special type (NST)(c), red bars: squamous cell carcinoma (SCC) (c), green bars: spindle cell carcinoma (SpCC)(c), purple bars: carcinoma with cartilaginous/osseous differentiation (CAR/OSS)(c), grey bars: matrix-producing carcinoma (MPC)(c), brown bars: pleomorphic carcinoma (PLM)(c).Percentages were calculated as the positive number/ total number of samples for each component.Significance was determined by the chi-square test, *P < 0.05 and **P < 0.01.n: total number of each sample /doi.org/10.1007/s12672-024-00950-0relationship between SCC(c) and PTEN loss has thus been demonstrated considering the entire SCC tumour mass.The present study is the first to reveal that PTEN loss and PTEN gene mutations are characteristically detected in SCC(c).

Table 1
Results of PTEN mutation analysis

Table 2
Results of next-generation sequencing analysis of 2 cases of MBC including SCC(c) Chr: chromosome; RefSeq: NCBI accession number; HGVS: human genome variant society; VAF: variant allele frequency; NT: not tested; -: not detected; IC: intraductal component; NST: no special type; SCC: squamous cell carcinoma Serial case no