The clinical relevance of apocrine morphology in breast cancer depends on the estrogen receptor status Authors: Kei Kawashima, Masanori Oshi, Akimitsu Yamada, Kazutaka Narui, Zunairah Shah, Thaer Khoury, Itaru Endo, Kazuaki Takabe Published: 2026 ABSTRACT AbstractApocrine morphology is a distinctive histopathological feature of breast cancer associated with unique clinical behavior, particularly in estrogen receptor (ER)-negative tumors. However, its molecular basis remains unclear and its relevance in ER-positive breast cancer has not been systematically examined. We performed an integrative analysis of breast cancer patients from the SEER database (n = 226,584) and The Cancer Genome Atlas (TCGA; n = 1017), incorporating clinicopathological, transcriptomic, and genomic data. Apocrine carcinoma was associated with favorable prognosis in ER-negative but worse survival in ER-positive breast cancer. Molecular and AR-related classifications did not fully recapitulate apocrine morphology. In ER-negative tumors, apocrine morphology correlated with lower proliferation and reduced predicted chemotherapy sensitivity. Conversely, in ER-positive tumors, apocrine morphology was associated with increased proliferation, enhanced immune-related signaling, and greater genomic instability. In conclusion, apocrine morphology represents an ER-dependent biological state with distinct prognostic and molecular implications.Subject terms: Cancer, Computational biology and bioinformatics, Oncology Introduction Apocrine morphology refers to the histologic features of apocrine glands, specialized androgen-responsive sweat glands that become active at puberty and secrete lipid-rich sweat, primarily located in the axilla1. The apocrine cells comprising these glands exhibit a distinctive morphology, including large, cuboidal or columnar cells with distinct cell borders, abundant granular eosinophilic or vacuolated cytoplasm, large round to oval nuclei, and prominent nucleoli2. Notably, apocrine morphology is not restricted to sweat glands but is also frequently observed in the breast through apocrine metaplasia of mammary epithelial cells3 with autopsy studies indicating that 50–85% of normal breasts exhibit some degree of apocrine change4–6. Apocrine carcinoma of the breast, a malignant breast tumor composed of apocrine-like epithelial cells, was first described in 1916 by Krompecher7, and was subsequently recognized as a distinct entity starting from the second edition of WHO classification8. Apocrine carcinoma has historically been defined based on morphological features, and in the 5th edition of the WHO Classification of Breast Tumours, breast carcinomas in which >90% of tumor cells exhibit apocrine morphology are classified as carcinoma with apocrine differentiation (CAD)9. In addition to this morphological definition, it is now recognized that these tumors are typically characterized by an ER-negative and AR-positive phenotype. In clinical practice, however, apocrine morphology is not restricted to tumors fulfilling these canonical features. Breast cancers with apocrine morphology are not always limited to the typical ER-negative and AR-positive phenotype and are often referred to as apocrine-like tumors10. Because large registry-based datasets, including the Surveillance, Epidemiology, and End Results (SEER) program, rely on pathology diagnoses rendered at individual institutions rather than centralized review, diagnostic criteria are not always applied uniformly11. As a result, such datasets likely encompass a heterogeneous group of breast cancers with apocrine features, including apocrine-like tumors. This heterogeneity is reflected in the reported presence of ER-positive cases among tumors classified as apocrine carcinoma, accounting for approximately 23–33% of cases, although this remains substantially lower than the frequency observed in non-apocrine breast cancers10,12. Although apocrine carcinoma accounts for only approximately 0.4–4% of all breast cancers2,13, it has attracted increasing attention because of its distinct clinical behavior and unique biological characteristics2,3,14. Historically, clinical investigation of apocrine carcinoma has focused on ER-negative disease, particularly those with triple-negative features, which have been associated with lower proliferative activity15 and more favorable survival outcomes13,16,17 despite worse response to chemotherapy compared with non-apocrine triple-negative breast cancer18–21. However, the underlying biological and molecular mechanisms remain unclear. Moreover, prior studies have focused predominantly on tumors exhibiting the typical ER-negative and AR-positive phenotypes, leaving the biological and clinical significance of apocrine morphology in ER-positive breast cancer largely unexplored. Beyond conventional pathological assessment, several efforts have attempted to define apocrine carcinoma using molecular approached in order to better capture its underlying biology13. Given that androgen receptor (AR)–related signaling is a characteristic feature of apocrine differentiation, subsets of breast cancers enriched for AR-related gene expression were defined as molecular apocrine carcinoma2. The original concept of the molecular apocrine carcinoma, initially proposed by Farmer et al.22, was later refined by Guedj et al. through unsupervised clustering based on the expression of 256 genes23, leading to the establishment of the Cartes d’Identité des Tumeurs program (CIT) classification and its molecular apocrine subtype. However, accumulating evidence suggests that molecular apocrine carcinomas are not fully concordant with apocrine carcinomas14, underscoring unresolved challenges in defining and interpreting apocrine differentiation in breast cancer. Similarly, the luminal androgen receptor (LAR) subtype identified in transcriptomic classifications of triple-negative breast cancer, such as the Lehmann classification, is characterized by AR signaling but does not consistently correspond to apocrine morphology2,24. Given this discordance, there is a clear need for integrative studies that bridge morphologic features with molecular characteristics. To date, no study has systematically interrogated the molecular and biological features of apocrine morphology as defined by histopathological criteria across large breast cancer cohorts. In this study, we leveraged pathology data from large breast cancer cohorts to perform a comprehensive, subtype-agnostic analysis of apocrine morphology as one component of the broader biological spectrum of breast cancer. By integrating clinical, pathological, and molecular data, we systematically examined the biological, genomic, and clinical relevance of apocrine morphology in breast cancer. Results Apocrine carcinoma is associated with favorable prognosis in ER-negative, but with worse survival in ER-positive breast cancers We first investigated whether apocrine carcinoma is associated with survival in ER-negative and ER-positive breast cancer in the large SEER cohort. Apocrine carcinoma was identified using the ICD-O-3 histology code (8401/3) for apocrine adenocarcinoma as recorded in the registry. Among the 1,259,968 patients with invasive breast carcinoma, apocrine carcinoma accounted for 2728 (0.22%) overall. When stratified by immunohistochemically determined ER-positivity, the proportion of apocrine carcinoma was significantly higher in ER-negative than in ER-positive breast cancer (0.84% vs. 0.08%, p < 0.001). As expected, apocrine carcinoma was associated with significantly better disease-specific survival (DSS) compared with non-apocrine carcinoma in ER-negative breast cancer (hazard ratio (HR) = 1.55, 95% confidence interval (CI) 1.38–1.74, p < 0.05), whereas overall survival (OS) showed no significant difference (HR = 0.96, 95% CI 0.89–1.03, p = 0.227, Fig. 1). In contrast, apocrine carcinoma was associated with significantly worse DSS and OS in ER-positive breast cancer (DSS: HR = 0.82, 95% CI 0.69–0.97, p < 0.05, OS: HR = 0.86, 95% CI 0.77–0.95, p < 0.05). Our results show that apocrine carcinoma is associated with better survival in ER-negative, whereas with worse survival in ER-positive breast cancer. Analyses in HER2-positive breast cancer showed worse disease-specific survival (HR = 1.56, 95% CI 1.08–2.26, p < 0.05) without a significant difference in overall survival (HR = 1.17, 95% CI 0.92–1.49, p = 0.210; Supplementary Fig 2A). Prognostic impact of apocrine carcinoma compared with non-apocrine carcinoma in the SEER cohort. Kaplan-Meier survival curves of disease-specific survival (DSS) and overall survival (OS) comparing apocrine and non-apocrine carcinomas in ER-negative and ER-positive patients from the SEER cohort. Orange line represents apocrine carcinoma, and green line represents non-apocrine carcinoma. Hazard ratios (HRs) with 95% confidence intervals (CIs) were estimated using Cox proportional hazards models, and log-rank test P values are shown. Numbers at risk are displayed below each plot. Apocrine-high tumors were associated with larger tumor size and lymph node metastasis regardless of ER-positivity To further characterize the clinical relevance of apocrine morphology, we next examined its association with clinicopathological features in the TCGA breast cancer cohort. We utilized pathological annotation data for apocrine morphology of TCGA cohort, and defined apocrine high as the breast cancers consist of more than 50% of apocrine morphology, and apocrine low as less than 50%. Clinicopathological characteristics of apocrine-high and apocrine-low tumors stratified by ER status are summarized in Table 1. The number of patients stratified in apocrine high and low groups were 84 (8.3%) and 933 (91.7%) respectively. When stratified by ER-positivity, apocrine-high was 20 out of 225 patients (8.9%) in ER-negative tumors, whereas 58 out of 746 patients (7.8%) in ER-positive tumors (p = 0.67). In ER-negative breast cancer, patients with apocrine-high tumors were significantly older than those with apocrine-low tumors (median age 63 vs. 54 years, p = 0.007). There were no significant differences in menopausal status, sex, race, histological grade, or HER2 status between the two groups. However, apocrine-high tumors were associated with more advanced pathological stage, with significant differences observed in both tumor size (p = 0.029) and nodal status (p = 0.012). Specifically, apocrine-high tumors showed a higher proportion of pT4 disease and nodal involvement compared with apocrine-low tumors. In ER-positive breast cancer, age, menopausal status, sex, and race did not differ significantly between apocrine-high and apocrine-low groups. In contrast, apocrine-high tumors were more frequently HER2-positive compared with apocrine-low tumors (35% vs. 20%, p = 0.010). While histological grades did not reach statistical significance (p = 0.069), apocrine-high tumors tended to show a higher proportion of grade 3 disease. Notably, apocrine-high ER-positive breast cancer was significantly associated with larger tumor size (p = 0.001) and with lymph node metastasis (p = 0.009). In addition, histological subtype distributions differed according to apocrine morphology in both ER-negative and ER-positive breast cancer. In ER-negative tumors, apocrine-high cases showed a higher proportion of invasive ductal carcinoma (IDC) and a lower proportion of invasive lobular carcinoma (ILC), although this did not reach statistical significance (p = 0.8). In contrast, in ER-positive breast cancer, histological subtype distributions were significantly different between apocrine-high and apocrine-low groups (p = 0.003), with apocrine-high tumors showing a higher proportion of IDC and a lower proportion of ILC. Similarly, PAM50 intrinsic subtype distributions were significantly associated with apocrine morphology in both ER-negative and ER-positive breast cancer (both p = 0.004). In ER-negative tumors, apocrine-high cases were relatively enriched for the HER2-enriched subtype and showed a reduced proportion of basal-like tumors. In ER-positive tumors, apocrine-high cases demonstrated a relative enrichment of HER2-enriched and luminal B subtypes, with a lower proportion of luminal A tumors compared with apocrine-low tumors. Table 1Clinicopathological characteristics according to apocrine morphology status in ER-negative and ER-positive breast cancersER-negativeER-positiveVariableOveralln = 225aLown = 205aHighn = 20ap valuebOveralln = 746aLown = 688aHighn = 58ap valuebAge55 [26–90]54 [26–90]63 [36–90]0.00760 [26–90]60 [26–90]59 [34–90]0.7MenopausePost153 (73%)140 (72%)13 (76%)>0.9506 (74%)465 (73%)41 (80%)0.3Pre58 (27%)54 (28%)4 (24%)180 (26%)170 (27%)10 (20%)Unknown1411360537RaceAsian19 (8.9%)18 (9.2%)1 (5.3%)0.334 (5.1%)31 (5.0%)3 (5.7%)0.9Black62 (29%)53 (27%)9 (47%)101 (15%)94 (15%)7 (13%)Other1 (0.5%)1 (0.5%)0 (0%)000White132 (62%)123 (63%)9 (47%)533 (80%)490 (80%)43 (81%)Unknown1110178735HER2Negative159 (79%)147 (81%)12 (63%)0.081536 (79%)505 (80%)31 (65%)0.01Positive42 (21%)35 (19%)7 (37%)141 (21%)124 (20%)17 (35%)Unknown24231695910Grade110 (4.5%)8 (4.0%)2 (10%)0.3193 (26%)185 (27%)8 (14%)0.069245 (20%)41 (20%)4 (20%)354 (48%)325 (47%)29 (51%)3167 (75%)153 (76%)14 (70%)195 (26%)175 (26%)20 (35%)Unknown330431pTT151 (23%)47 (23%)4 (20%)0.029195 (26%)184 (27%)11 (19%)0.001T2144 (64%)133 (65%)11 (55%)418 (56%)385 (56%)33 (57%)T319 (8.5%)18 (8.8%)1 (5.0%)107 (14%)101 (15%)6 (10%)T410 (4.5%)6 (2.9%)4 (20%)25 (3.4%)17 (2.5%)8 (14%)Unknown110110pNN0129 (58%)122 (60%)7 (37%)0.012329 (45%)306 (45%)23 (43%)0.009N159 (26%)55 (27%)4 (21%)264 (36%)251 (37%)13 (25%)N220 (9.0%)16 (7.8%)4 (21%)84 (11%)70 (10%)14 (26%)N315 (6.7%)11 (5.4%)4 (21%)54 (7.4%)51 (7.5%)3 (5.7%)Unknown21115105Histology0.80.003IDC197 (87%)178 (86%)19 (95%)520 (69%)468 (67%)52 (90%)ILC7 (3.1%)7 (3.4%)0 (0%)169 (22%)165 (24%)4 (6.9%)Mucinous0 (0%)0 (0%)0 (0%)14 (1.8%)14 (2.0%)0 (0%)Other/Unknown22 (9.7%)21 (10%)1 (5.0%)55 (7.3%)53 (7.6%)2 (3.4%)PAM500.0040.004Normal11 (5.3%)8 (4.2%)3 (17%)18 (2.7%)18 (2.9%)0 (0%)Basal139 (67%)133 (70%)6 (33%)20 (3.0%)18 (2.9%)2 (3.6%)LumA11 (5.3%)11 (5.8%)0 (0%)432 (64%)406 (66%)26 (46%)LumB3 (1.4%)3 (1.6%)0 (0%)176 (26%)154 (25%)22 (39%)Her245 (22%)36 (19%)9 (50%)26 (3.9%)20 (3.2%)6 (11%)Unknown1715286842aData are presented as median [range] or number (%).bP values were calculated using the Wilcoxon rank-sum test, Pearson’s chi-squared test, or Fisher’s exact test as appropriate. Bold values indicate statistical significance (p < 0.05).Patients were stratified by estrogen receptor (ER) status and further classified according to apocrine morphology (high vs. low). Continuous variables are presented as median [range], and categorical variables as number (%). P values were calculated using the Wilcoxon rank-sum test for continuous variables and Fisher’s exact test or Pearson’s chi-squared test for categorical variables, as appropriate.IDC Invasive ductal carcinoma, ILC Invasive lobular carcinoma. Clinicopathological characteristics according to apocrine morphology status in ER-negative and ER-positive breast cancers aData are presented as median [range] or number (%). bP values were calculated using the Wilcoxon rank-sum test, Pearson’s chi-squared test, or Fisher’s exact test as appropriate. Bold values indicate statistical significance (p < 0.05). Patients were stratified by estrogen receptor (ER) status and further classified according to apocrine morphology (high vs. low). Continuous variables are presented as median [range], and categorical variables as number (%). P values were calculated using the Wilcoxon rank-sum test for continuous variables and Fisher’s exact test or Pearson’s chi-squared test for categorical variables, as appropriate. IDC Invasive ductal carcinoma, ILC Invasive lobular carcinoma. Apocrine morphology is associated with ER signaling and is not fully captured by AR-related features or molecular apocrine classification Given that the clinical relevance of apocrine morphology differed markedly by ER status, we examined ER-related transcriptional programs in TCGA cohort. Neither ESR1 (gene encoding ER-alpha) nor ESR2 (gene encoding ER-beta) expression levels were significantly correlated with apocrine morphology in either ER-negative (ESR1: p = 0.407; ESR2: p = 0.179) or ER-positive breast cancer (ESR1: p = 0.648; ESR2: p = 0.087; Fig. 2D left). In contrast, gene set variation analysis of the Hallmark estrogen response pathways (early and late) revealed opposite patterns according to ER status: in ER-negative breast cancer, apocrine-high tumors exhibited significantly higher estrogen response scores, whereas in ER-positive breast cancer, apocrine-high tumors showed significantly lower estrogen response activity (all p < 0.05; Fig. 2D right). These results indicate that apocrine morphology relates to differential regulation of estrogen-responsive gene programs according to ER status, rather than simply reflecting receptor abundance. Histologic features and hormone-related molecular characteristics of apocrine morphology in breast cancer. A–C Representative H&E-stained slides of apocrine-high tumors and apocrine-low tumors. A Invasive carcinoma with pure apocrine morphology showing classic features including conspicuous eosinophilic cytoplasm, enlarged nuclei (some cells show more than one nucleus), and prominent nucleoli; B and C from the same tumor, where B show apocrine component ( < 50% in this case), and (C) non-apocrine (all images Hematoxylin and Eosin stain, 20× magnification). D–G Association of apocrine morphology with estrogen- and androgen-related features, and molecular subtypes in the TCGA cohort. D Box plots showing ESR1 and ESR2 expression levels and GSVA scores for the MSigDB Hallmark estrogen response early and late gene sets, stratified by apocrine morphology-high and -low group within ER-negative and ER-positive breast cancers (E) Bar plots of AR mutation frequency and AR amplification frequency, and box plots of inferred AR transcription factor activity, AR gene-expression level and GSVA scores for the MSigDB Hallmark androgen response gene set, stratified by apocrine morphology-high and -low within ER-negative and ER-positive tumors in the TCGA cohort. In all box plots, the center line indicates the median, the box represents the interquartile range (IQR), and whiskers extend to 1.5× IQR. Outliers were omitted. P values were calculated using the Wilcoxon rank-sum test. F Pie chart showing the distribution of CIT classification (basal, basL; luminal-A, lumA; luminal-B, lumB; luminal-C, lumC; molecular apocrine, mApo; normal, normL) (G) Stacked bar plot showing the distribution of apocrine-low and apocrine-high tumors across CIT classification. Percentages within each subtype are indicated within each bar. Since previous reports linked apocrine morphology with AR signaling, we next investigated the association of AR-related molecular features and apocrine morphology. To investigate whether genomic alterations in AR contribute to these associations, we first examined AR mutation and copy number alteration profiles. AR mutations were rare in both ER-negative and ER-positive breast cancer and showed no significant differences between apocrine-high and apocrine-low tumors in either subgroup, although a trend toward lower mutation frequency in apocrine-high tumors was observed. Similarly, AR amplification tended to be more frequent in apocrine-high tumors; however, these differences did not reach statistical significance. We next assessed AR gene expression and transcription factor activity. In ER-negative breast cancer, apocrine-high tumors exhibited significantly higher AR transcription factor activity and AR gene expression compared with apocrine-low tumors (all p < 0.05, Fig. 2E). In contrast, in ER-positive breast cancer, no significant differences were observed in AR transcription factor activity or AR gene expression between apocrine-high and apocrine-low groups (p = 0.341 and p = 0.181, respectively). Finally, gene set variation analysis of the MSigDB Hallmark androgen response pathway revealed that androgen response signaling was significantly enriched in apocrine-high tumors in both ER-negative and ER-positive breast cancer (all p < 0.05, Fig. 2E). These findings imply that in ER-positive disease, enhanced androgen response signaling associated with apocrine morphology may not be directly driven only by genomic alterations or increased AR expression or transcriptional activity, particularly in ER-positive disease, and may instead reflect alternative regulatory mechanisms influencing downstream androgen-responsive gene programs. In HER2-positive breast cancer, apocrine morphology showed no significant association with ER-related genes or estrogen response pathways, whereas significant associations were observed with AR-related molecular features (Supplementary Fig S2). Next, we applied the CIT classification to the TCGA cohort using the CITBCMST package. The molecular apocrine (mApo) subtype, defined by a characteristic gene-expression profile that includes AR-related signaling, accounted for 7% of all breast cancer patients in TCGA (Fig. 2F). When examined apocrine morphology across CIT classification, although the mApo subtype showed the highest proportion of apocrine-high tumors among molecular subtypes (24%), the majority of mApo cases still fell into the apocrine-low category (Fig. 2G), indicating that molecular subtyping alone does not fully capture apocrine morphology. Apocrine morphology is associated with distinct mutational landscapes by the ER status in breast cancer Given that apocrine morphology in the breast commonly arises in the setting of apocrine metaplasia, we sought to investigate the mutational landscape of these tumors. In ER-negative breast cancer, no consistent differences were observed between apocrine-high and apocrine-low tumors across genomic instability metrics. In contrast, within ER-positive breast cancer, the apocrine-high group exhibited significantly greater intratumor heterogeneity, altered fraction HRD, silent and non-silent mutation burden, and SNV-derived neoantigen load, indicating increased genomic instability and a more complex mutational landscape (Fig. 3A). At the genomic level, apocrine morphology showed opposite associations with the mutation status of PIK3CA, a canonical luminal driver and one of the most frequently altered genes in ER-positive breast cancer, depending on ER context. In ER-negative tumors, PIK3CA mutations were more frequent in the apocrine-high group (28%), ranking as the third most common alteration, compared with 14% in apocrine-low tumors (p < 0.05). Conversely, in ER-positive tumors, PIK3CA was the most frequently mutated gene in the apocrine-low group (51%), consistent with its canonical role in luminal breast cancer, but occurred significantly less often in apocrine-high tumors (28%, p < 0.05; Fig. 3B). In HER2-positive tumors, apocrine morphology showed no significant association with genomic instability indices or with the frequency of individual gene mutations (Supplementary Fig S3). Association of apocrine morphology with tumor immunogenicity-related scores and mutational landscapes. A Box plots showing intratumor heterogeneity, altered fraction, homologous recombination deficiency (HRD) score, non-silent and silent mutation burdens, and neoantigen load (SNV- and indel-derived) stratified by apocrine morphology across subtypes. B Oncoplots depicting recurrent somatic alterations in ER-negative and ER-positive tumors, stratified by apocrine morphology. Alteration frequencies are shown on the right of each panel. Collectively, these findings indicate that apocrine morphology is associated with distinct, ER-dependent mutational patterns—characterized by increased genomic instability in ER-positive disease and a context-dependent relationship with PIK3CA mutations in ER-negative versus ER-positive breast cancer. Apocrine morphology is associated with low cell proliferation in ER-negative breast cancer and with high cell proliferation in ER-positive breast cancer Since the association of apocrine morphology and genomic instability as well as survival were the opposite by the ER status, we next investigated the differences in proliferative activities according to the extent of apocrine morphology. Gene set enrichment analysis (GSEA) revealed a clear association of cell proliferation-related pathways by the ER status (Fig. 4A). In ER-negative breast cancer, apocrine-high tumors showed significant suppression of cell proliferation-related gene sets, including G2M checkpoint, E2F targets, MYC targets (v1/v2), and mitotic spindle (all FDR < 0.25). In contrast, in ER-positive breast cancer, G2M checkpoint, E2F targets, MYC targets (v1), and mitotic spindle were significantly enriched in apocrine-high tumors (FDR < 0.25), whereas MYC targets (v2) did not reach statistical significance (FDR = 0.775), indicating opposite regulation of proliferative programs according to ER status. Consistent with the GSEA results, in ER-negative breast cancer, apocrine-high tumors demonstrated significantly lower proliferation scores and lower Ki67 gene expression compared with apocrine-low tumors (both p < 0.05; Fig. 4B). Further, the proportion of apocrine-high tumors was highest among histological grade 1 cases, although this association was not statistically significant (p = 0.46). In contrast, in ER-positive breast cancer, apocrine-high tumors exhibited significantly higher proliferation scores calculated by Thorsson et al. (p < 0.05). The percentage of apocrine-high tumors trended to be associated with higher Nottingham histological grades, and Ki67 gene expression was also higher compared with apocrine-low tumors, although these differences did not reach statistical significance (grade, p = 0.0687; Ki67, p = 0.124). In HER2-positive tumors, apocrine-high cancers exhibited significantly lower MKI67 expression, but no significant differences were detected in proliferation-related pathways by GSEA (Supplementary Fig. S4). Association of apocrine morphology with proliferative activity. A Gene set enrichment analysis (GSEA) enrichment plots for proliferation-related MSigDB Hallmark pathways (G2M checkpoint, E2F targets, MYC targets v1/v2) across ER-negative and ER-positive tumors. Normalized enrichment score (NES) and false discovery rate (FDR) are shown for each pathway. B Box plots comparing Ki-67 expression between apocrine-high and -low tumors (left), bar plots showing distribution of apocrine-high tumor across tumor grade (middle), and box plots comparing proliferation score between apocrine-high and -low tumors (right) across ER-negative and ER-positive tumors in the TCGA cohort. P values were calculated using the Wilcoxon test. Apocrine morphology is associated with reduced immune activity in ER-negative breast cancer but with increased immune activity in ER-positive disease Next, we investigated differences in immune activity according to apocrine morphology stratified by ER status. GSEA using the MSigDB Hallmark gene sets revealed a clear difference by the ER status (Fig. 5A). In ER-negative breast cancer, interferon-alpha and -gamma response as well as allograft rejection were consistently suppressed (all FDR < 0.25), while TGF-beta signaling showed an opposite trend (FDR < 0.25). In contrast, in ER-positive breast cancer, apocrine-high tumors demonstrated significant enrichment of interferon-alpha and -gamma response (both FDR < 0.25), whereas TGF-beta signaling exhibited an opposite but non-significant trend (FDR = 0.871). These pathway-level findings were further examined using additional immune metrics. Overall, apocrine morphology was not significantly associated with marked differences in immune infiltration metrics. In ER-negative breast cancer, apocrine morphology was not associated with significant differences in histology-based TIL scores (p = 0.498), immune scores (p = 0.800), or CYT scores (p = 0.112) despite a general trend toward lower immune activity in apocrine-high tumors. Similarly, in ER-positive breast cancer, no significant differences were observed in TIL scores (p = 0.157) or CYT scores (p = 0.325) between apocrine-high and apocrine-low tumors. Although the xCell-derived immune score was modestly higher in apocrine-high tumors, this represented the only immune metric showing statistical significance (p < 0.05). In HER2-positive tumors, apocrine-high cancers were associated with enrichment of immune-related pathways by GSEA and high CYT scores (Supplementary Fig S5). Association of apocrine morphology with immune-related activity and immune-associated scores. A Gene set enrichment analysis (GSEA) enrichment plots for immune-related Hallmark gene sets (interferon alpha signaling, interferon gamma signaling, allograft rejection, and TGF-β signaling) comparing apocrine-high and apocrine-low tumors. Normalized enrichment score (NES) and false discovery rate (FDR) are shown for each pathway. B Box plots showing tumor-infiltrating lymphocyte (TIL) score (left), immune score (middle), and cytolytic activity (CYT) score (right) stratified by apocrine morphology-high and -low group. P values for box plots were calculated using the Wilcoxon test. Higher predicted chemotherapy sensitivity in apocrine-low tumors in ER-negative breast cancer As apocrine morphology was associated with subtype-specific differences in immune-related pathway activity, we next sought to determine whether these biological differences translated into differential therapeutic vulnerability. To this end, we evaluated predicted drug response using the oncoPredict package, which estimates tumor-specific drug sensitivity based on transcriptomic profiles and reference drug response data from cancer cell lines (Fig. 6). In the resulting sensitivity scores, lower values reflect greater predicted sensitivity to the corresponding agent. In ER-negative breast cancer, apocrine-low tumors exhibited significantly lower predicted drug response values, indicating greater predicted sensitivity to 5-FU, cisplatin, paclitaxel, olaparib, and talazoparib (all p < 0.05), compared with apocrine-high tumors. In contrast, in ER-positive breast cancer, no significant differences in predicted sensitivity were observed between apocrine-high and apocrine-low tumors across commonly used therapeutic agents, despite marked differences in proliferation and immune activity. Similarly, in HER2-positive breast cancer, apocrine morphology was not significantly associated with predicted sensitivity to any of the therapeutic agents (Supplementary Fig S6). Association of apocrine morphology with predicted response to anticancer therapeutics. Box plots comparing predicted drug response scores for standard chemotherapeutic agents and targeted therapies between apocrine-marked and apocrine-negative tumors across ER-negative and ER-positive tumors. Lower values reflect greater predicted sensitivity to the corresponding agent. P values were calculated using the Wilcoxon test. Discussion This study identifies apocrine morphology as a unique aspect of breast cancer biology that is not fully captured by the AR expression-based molecular apocrine subtype. Although ER-negativity was long considered one of the diagnostic criteria for carcinoma with apocrine differentiation, we found that clinical and biological relevance of apocrine morphology are opposite by the ER status. In ER-negative, apocrine carcinoma was associated with more favorable clinical outcomes compared with non-apocrine carcinoma. ER-negative breast cancer with high apocrine morphology exhibited lower proliferative activity and lower predictive sensitivity to chemotherapy, in line with previous reports13,15–19. In contrast in ER-positive breast cancer, apocrine carcinoma was associated with poorer prognosis, and apocrine morphology was associated with higher proliferative activity, increased mutation burden, and enriched immune-related features, representing an opposite phenotype to that observed in ER-negative breast cancer. Together, these findings suggest that apocrine morphology may serve as an ER-dependent indicator with potential utility for prognostic stratification and therapeutic decision-making in breast cancer. One of the key features closely associated with apocrine morphology is AR signaling, as apocrine cells retain phenotypic characteristics of apocrine glands, which are regulated in an androgen-dependent manner25. However, although immunohistochemical studies have reported significantly higher AR expression in apocrine carcinoma compared with conventional TNBC26, and AR positivity has been associated with a higher frequency of apocrine morphology in ER-negative tumors27–30, AR expression did not uniformly correspond to morphologically defined apocrine carcinoma even in these studies. Indeed, among AR-positive ER-negative breast cancers, only about a quarter have been reported to exhibit apocrine morphology29, underscoring the incomplete concordance between AR expression and apocrine morphology. Moreover, the absence of such an association in ER-positive cases has also been described27, suggesting that the relationship between AR expression and apocrine morphology may differ according to ER status. AR splice variants, particularly AR-V7, may provide one possible explanation for this discordance. AR-V7 encodes a truncated receptor lacking the ligand-binding domain and exhibits constitutive, ligand-independent transcriptional activity. Previous studies have reported that apocrine morphology is more frequently observed in AR-V7–positive breast cancers compared with AR-V7–negative cases, even within AR IHC–positive tumors31,32. This suggests that AR splice variants may contribute to apocrine differentiation beyond AR expression alone. Moreover, despite some claim the classification of molecular apocrine carcinoma based on AR-related gene-expression profiles22,23,33; only 7–24% of molecular apocrine tumors have been reported to exhibit apocrine morphology14,29. In our study, apocrine morphology in ER-negative breast cancer showed significant correlations with AR gene expression, AR-associated transcription factors, and androgen response pathway activity, while AR genomic alterations, including copy number variations and mutations, showed a modest but consistent association. In contrast, in ER-positive breast cancer, apocrine morphology was associated with enrichment of androgen response pathways, whereas AR genomic alterations, including copy number variations and mutations, were not significantly associated with apocrine morphology, and neither AR gene expression nor AR-related transcription factors demonstrated a significant correlation. This dissociation between AR gene expression and androgen-responsive transcriptional activity in ER-positive tumors suggests that apocrine morphology may reflect activation of androgen-related transcriptional programs that are not solely dependent on AR expression levels. In addition, it has been suggested that the biological effects of AR signaling differ depending on ER status. In the absence of ER, AR may function as an alternative transcriptional driver, whereas in ER-positive tumors, AR signaling operates within ER-driven transcriptional networks and may exert competitive or cooperative effects34. This context-dependent role of AR may represent one potential explanation for the opposing biological associations of apocrine morphology observed between ER-positive and ER-negative breast cancers. Consistent with this interpretation, analysis based on CIT classification demonstrated that although the proportion of apocrine-high tumors was highest within the molecular apocrine subtype, these tumors accounted for only 24% of molecular apocrine cases, with the majority of molecular apocrine tumors exhibiting low apocrine morphology, underscoring a marked discordance between molecular and morphological classifications. Collectively, these findings indicate that although apocrine morphology is biologically linked to androgen-related programs, molecular AR-associated features cannot fully substitute for morphologic classification. Previous studies have consistently suggested that apocrine carcinoma in ER-negative breast cancer is associated with a more favorable prognosis compared with non-apocrine counterparts13,16,17,35. In our analysis of the SEER cohort, apocrine carcinoma was likewise associated with favorable outcomes in ER-negative breast cancer, consistent with prior reports. From an immunological perspective, prior studies have also reported higher levels of TILs in apocrine tumors lacking ER expression compared with those with ER expression36, potentially contributing to these favorable outcomes. However, the biological mechanisms underlying those prognostic differences have remained poorly identified. In this study, apocrine morphology was significantly associated with lower proliferative activity in ER-negative tumors, providing a potential biological basis for favorable prognosis in these subtypes. In contrast, the prognostic significance in ER-positive breast cancer has not been systematically examined. Notably, we found that apocrine carcinoma in ER-positive breast cancer was associated with worse clinical outcomes compared with non-apocrine tumors. Concordantly, apocrine morphology in ER-positive breast cancer showed a positive correlation with proliferative activity, along with enrichment of immune-related pathways. However, no statistically significant differences were observed in TIL or cytolytic activity scores in either ER-positive or ER-negative tumors, although trends were broadly consistent with pathway-level findings, suggesting that these immune-related changes may be modest or not fully captured by these measures. Together, these features are indicative of a more aggressive tumor biology and suggest that the biological and clinical impact of apocrine morphology is highly ER-dependent. These ER status-specific biological states in apocrine carcinoma were further supported by distinct genomic features associated with apocrine morphology. In HER2-positive breast cancer, previous reports have suggested a potentially favorable prognosis for apocrine carcinoma35, although detailed investigations remain limited. In our exploratory analyses, HER2-positive apocrine tumors were likewise associated with improved clinical outcomes compared with non-apocrine tumors and showed an inverse association with proliferative indices. In ER-positive breast cancer, PIK3CA is a canonical luminal driver mutation and one of the most frequently mutated genes, and has been reported to be associated with favorable prognosis37,38. However, in this study, apocrine-high tumors showed a significantly lower frequency of PIK3CA mutations, a pattern that contrasts with the established role of PIK3CA in ER-positive disease and has never been previously shown in the context of apocrine morphology. Instead, these tumors were characterized by higher homologous recombination deficiency, increased mutation burden, and greater intratumor heterogeneity, indicating a shift from canonical luminal oncogenic pathways toward genome instability–driven mechanisms. In contrast in ER-negative breast cancer, apocrine morphology was associated with a higher frequency of PIK3CA mutations, consistent with previous reports24,39–41, highlighting a context-dependent role of PIK3CA alterations across breast cancer subtypes. This study has several limitations that should be acknowledged. First, in the SEER cohort, apocrine carcinoma was identified using the ICD-O-3 histology code recorded in the registry. Because this coding reflects routine diagnostic classification in clinical practice, it does not rely on centralized pathology review, and detailed morphologic reassessment was not feasible. Second, in the TCGA cohort we defined apocrine-high tumors as those with >50% apocrine morphology, which does not strictly correspond to the WHO diagnostic criterion requiring >90% of tumor cells to exhibit apocrine differentiation. Third, predicted chemotherapy sensitivity was estimated using the oncoPredict framework, an in-silico model based on tumor transcriptomic profiles that does not directly measure clinical treatment response. Finally, this study was observational in nature, and the limited availability of datasets with detailed, standardized annotation of apocrine morphology precluded functional validation and extensive external validation in independent cohorts. Future studies incorporating centralized pathological review and prospective validation are warranted to further clarify the biological and clinical significance of apocrine morphology. This study provides a comprehensive analysis of apocrine morphology in breast cancer stratified by ER status. In ER-negative disease, apocrine carcinoma was associated with a favorable prognosis, and apocrine morphology correlated with lower proliferative activity. In contrast, in ER-positive breast cancer, apocrine carcinoma was associated with adverse outcomes, with apocrine morphology correlating with increased proliferation, enhanced immune signaling, and greater genomic instability. Collectively, these findings indicate that apocrine morphology represents an ER-dependent biological state and may serve as a biomarker with potential utility for prognostic stratification and therapeutic decision-making in breast cancer. Methods Data acquisition Breast cancer patient data from the Surveillance, Epidemiology, and End Results (SEER) Program database (n = 1,259,968) and The Cancer Genome Atlas (TCGA) (n = 1,017) were analyzed in this study. For the SEER cohort, clinical and pathological data were obtained from the “Incidence—SEER Research Data, 17 Registries (2000–2022)” using SEER*Stat software. Patients with primary invasive epithelial breast cancer (ICD-O-3 site codes C50.0-C50.9; histology codes 8000-8576) with documented immunohistochemically assessed ER status were included. Apocrine carcinoma cases were identified based on a pathological diagnosis of invasive apocrine carcinoma registered in the database (ICD-O-3 code 8401/3). Subsequent analyses were performed using Surveillance Research Program, National Cancer Institute SEER*Stat software (seer.cancer.gov/seerstat) version 9.0.42.2. For the TCGA cohort, clinical, transcriptomic and whole exome sequencing (WES) data were downloaded from cBioPortal and the UCSC Xena platform. Apocrine morphology annotations for the TCGA cohort were derived from the dataset reported by Thennaven et al.42, which was based on systematic assessment of virtual hematoxylin and eosin-stained (H&E) whole-slide images. In this dataset, apocrine morphology was quantified according to the proportion of invasive carcinoma exhibiting apocrine features and categorized as Absent, Minimal (1–5%), Moderate (6–50%), or Marked (>50%). Based on this classification, we defined tumors with marked apocrine morphology (>50%) as apocrine-high, whereas those with absent, minimal, or moderate apocrine features (≤50%) as apocrine-low group. Representative H&E images from each apocrine morphology group are shown in Fig. 2A–C. In subsequent analyses, tumors were stratified by ER positivity, as defined by TCGA clinical annotations, and comparisons between apocrine-high and apocrine-low groups were performed separately within ER-positive and ER-negative subsets. In addition, exploratory analyses were conducted for HER2-positive tumors, and the corresponding results are presented in Supplementary Figs. S1–S6. All datasets were accessed and downloaded in December 2025. Because all data used in this study were obtained from publicly available, de-identified sources, institutional review board (IRB) approval was not required. Pathway and regulatory network analysis Differentially expressed genes (DEGs) were identified using the limma package in R. Gene set enrichment analysis (GSEA) was subsequently performed using the fgsea package with the Hallmark gene set collection from MSigDB43. Genes were ranked by log₂ fold-change values, and GSEA was conducted in a pre-ranked manner. Statistical significance was defined as a false discovery rate (FDR) < 0.25, and enrichment magnitude was assessed using the normalized enrichment score (NES). In addition, transcription factor (TF) activity was inferred using the DoRothEA package, which estimates TF activity based on the expression of curated target gene regulons. Molecular subtyping Molecular subtyping of TCGA breast cancer samples was conducted using the CITBCMST package in R44,45, which implements the CIT classification23. The CIT classification is a clustering-based approach that defines six stable molecular subtypes based on the expression of 256 genes. In this framework, tumors are grouped according to gene-expression patterns, and each resulting cluster is labeled as a molecular subtype. The CIT classification refines previous molecular subtyping frameworks by integrating hormone receptor signaling and proliferative activity. Specifically, tumors are first stratified by ER-related gene expression, distinguishing ER-positive and ER-negative groups. Among ER-positive tumors, subtypes are further separated based on ER expression levels and proliferation-associated gene expression, resulting in Normal-like (NormL), Luminal-A (LumA), Luminal-B (LumB), and Luminal-C (LumC) subtypes. In contrast, ER-negative tumors are divided according to AR expression, with AR-positive tumors classified as molecular apocrine (mApo) and AR-negative tumors classified as basal-like (basL).According to CIT classification, tumors were assigned to one of the following subtypes: basL, lumA, lumB, lumC, mApo, and NormL. Somatic mutation analysis Somatic mutation data were analyzed using Mutation Annotation Format (MAF) files. Nonsynonymous somatic mutations were retained, and oncoplots were generated using the maftools package in R. Immunologic and genomic features Immune cell infiltration within the tumor microenvironment (TME) was estimated using the xCell algorithm46, and cytolytic activity (CYT) scores were calculated as the geometric mean of granzyme A (GZMA) and perforin (PRF1) expression as described by Rooney et al.47. Histology-based tumor-infiltrating lymphocyte (TIL) scores were obtained from the dataset reported by Cha et al.48, which was generated through annotation of virtual H&E–stained whole-slide images. Tumor immunogenicity and proliferative activity were assessed using pre-calculated scores reported by Thorsson et al.49 as our group has previously described50,51. Drug response prediction Clinical drug response to anticancer agents was predicted using the oncoPredict package52, which applies transcriptome-based modeling trained on cancer cell line drug sensitivity data (GDSC2) to generate predicted drug response values for each tumor sample. Lower predicted IC50 values indicate greater drug sensitivity. Statistical analysis Primary analyses were stratified by ER status to account for biological heterogeneity. Additional exploratory analyses were conducted in the HER2-positive subgroup as a supplementary assessment. Statistical analyses and data visualization were performed using R (version 4.5.1) with Bioconductor (version 3.21). Group comparisons were performed using the Kruskal-Wallis test, Mann-Whitney U test, or Fisher’s exact test, as appropriate. Survival analyses were conducted using Kaplan-Meier analysis with the log-rank test. A p value < 0.05 was considered statistically significant for all analyses. Supplementary information Apocrine_morphology_supplementary_figure_revision2_20260513_KK Acknowledgements K.T. is supported by US National Institutes of Health grant number R01CA287731, R01CA250412, R37CA248018, as well as US Department of Defense BCRP, grant number W81XWH-19-1-0674 and HT9425-24-1-0762. Roswell Park Comprehensive Cancer Center is supported by US National Institute of Health Cancer Center Support Grant number P30CA016056. Author contributions K.K. conceptualized and designed the study, performed statistical and bioinformatic analyses, and wrote the original draft. K.T. provided supervision, contributed to study design, and critically revised the manuscript. M.O., A.Y., K.N., Z.S., T.K., and I.E. critically revised the manuscript. All authors approved the final manuscript. Data availability All data analyzed in this study are publicly available. The TCGA breast cancer datasets were obtained from the cBioPortal (https://www.cbioportal.org/) and the UCSC Xena platform (https://xenabrowser.net/). Apocrine morphology annotations for TCGA were obtained from the publicly available dataset reported by Thennavan et al 42. The SEER data are available through the SEER*Stat software (https://seer.cancer.gov/) upon registration. No new datasets were generated during the current study. The code used for the analyses in this study is available from the corresponding author upon reasonable request. Competing interests The authors declare no competing interests. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Supplementary information The online version contains supplementary material available at https://doi.org/10.1038/s41523-026-00981-6. References Associated Data Supplementary Materials Apocrine_morphology_supplementary_figure_revision2_20260513_KK Data Availability Statement All data analyzed in this study are publicly available. The TCGA breast cancer datasets were obtained from the cBioPortal (https://www.cbioportal.org/) and the UCSC Xena platform (https://xenabrowser.net/). Apocrine morphology annotations for TCGA were obtained from the publicly available dataset reported by Thennavan et al 42. The SEER data are available through the SEER*Stat software (https://seer.cancer.gov/) upon registration. No new datasets were generated during the current study. The code used for the analyses in this study is available from the corresponding author upon reasonable request.