Genomic landscape, immune microenvironment and survival in male versus female breast cancer.

Introduction
Breast cancer (BC) is a leading cause of morbidity and mortality in women, but it is rare in men: out of an estimated 20 000 new BC cases diagnosed in 2020 (2650 of them in the United States) only around 1% occurred in men.1,2 Identified risk-increasing conditions for male breast cancer (MaBC) include unbalanced lifetime estrogen exposure for such conditions as hereditary gonadal dysgenesis (e.g. Klinefelter syndrome) and possibly female gender-affirmation hormone treatments in transgender persons, family history of breast cancer, and germline cancer-predisposing mutations, mostly BRCA1 and BRCA2.3,4 In unselected MaBC, around 10% of patients harbor a pathogenetic germline mutation of BRCA2, with broad variability across the series in literature, whereas BRCA1 is described on average in 1% of patients.3,4 By contrast, ∼5%-10% of female breast cancer (FeBC) results from inherited mutations of BRCA genes. Additional germline mutations associated with MaBC include CHEK2, ATM, NF1, and PALB2; however, these mutations contribute individually for <2% of MaBC.
A cancer genomic characterization of MaBC has been described previously in a cohort of 59 MaBCs, evaluated with immunohistochemistry (IHC) and gene sequencing.5 MaBC presented almost exclusively as hormone receptor (HR)-positive and human epidermal growth factor receptor 2 (HER2)-negative ductal-type BC. The genomic landscape of MaBC displays key differences as compared with FeBC, including fewer mutations in PIK3CA and TP53, associated with endocrine resistance and poorer cancer prognosis in BC. MaBC also appeared enriched for mutations affecting DNA repair-related genes, including homologous recombination deficiency (HRD), recapitulating a key aspect of the MaBC tumorigenesis. Previous studies have consisted mostly of HR-positive/HER2-negative, yet HER2-positive BC in men accounts for ∼13% of all MaBC cases.6 So far, no comprehensive genomic and transcriptomic analysis of MaBC has been reported on a larger cohort, especially across relevant BC subtypes, and the biology of HR-negative or HER2-positive MaBC is largely unknown. The aim of our study is to describe the sex-based differences in genomic, transcriptomic, and immunomic characteristics of BC, exploring potential therapeutic implications and impact on real-world patient outcomes.

Materials and methods
A total of 39 601 formalin-fixed paraffin embedded (FFPE) BC samples were obtained from Caris Life Sciences (Phoenix, AZ). Samples were sequenced with next-generation sequencing (NGS)7 and whole-transcriptome sequencing (WTS).7,8 Immune signatures and cell infiltration were evaluated via RNA deconvolution,9 and T-cell-inflamed and interferon-gamma scores were calculated using validated panels.10,11 Tumor mutational burden (TMB) was defined as TMB-high at ≥10 mutations/MB.12 Statistical analyses, including Chi-square, Fisher Exact tests, and Spearman’s correlation, identified significant differences (P < 0.05), with corrections for multiple comparisons using the Benjamini–Hochberg method (q < 0.05). Real-world survival was analyzed using Kaplan–Meier curves.13,14

Ethics approval
This study was conducted in accordance with guidelines of the Declaration of Helsinki, Belmont report, and United States Common Rule. In keeping with 45 CFR 46.101(b)(4), this study was carried out utilizing retrospective, deidentified clinical data; sex was classified as reported in the medical records. This study is considered IRB exempt, and no patient consent was necessary from the subject.

Next-generation sequencing
NGS was carried out on genomic DNA using the NextSeq (592-whole gene targets; Agilent Technologies, Santa Clara, CA) or NovaSeq 6000 (>700 genes at high coverage/read-depth and >20 000 genes at lower depth; Agilent Technologies, Santa Clara, CA) platforms (Illumina, Inc., San Diego, CA). ‘Pathogenic’, and ‘likely pathogenic’ variants were detected as described previously.7

Whole-transcriptome sequencing
WTS was carried out using the Illumina Novaseq 6500 platform (Illumina, Inc., San Diego, CA) as described.7 Transcripts per million (TPM) molecules were generated using the Salmon expression pipeline.8 Differentially expressed genes were determined by calculating log2 fold-change between groups.

Tumor immune signature and immune cell infiltration
Immune gene signatures were determined through WTS expression profiles as described. Immune cell infiltration was carried out using RNA-deconvolution techniques via quanTIseq (Innsbruck, Austria).9 T-cell-inflamed scores were calculated using T-cell-inflamed gene expression signature consisting of 160 genes, as previously described.10 Interferon-gamma (IFN-γ) scores were calculated with a validated 18-gene panel [CL5, CD27, CD274 (PD-L1), CD276 (B7-H3), CD8A, CMKLR1, CXCL9, CXCR6, HLA-DQA1, HLA-DRB1, HLA-E, IDO1, LAG3, NKG7, PDCD1LG2 (PDL2), PSMB10, STAT1, and TIGIT].11

Tumor mutational burden
TMB was measured from NGS data as previously described (TMB-high ≥10 mutations/MB).12 Mismatch repair deficiency (dMMR)/microsatellite instability (MSI) status was determined by IHC and NGS as described previously.11,15

Immunohistochemistry
IHC of programmed death-ligand 1 (PD-L1; 22c3 and SP142 clone) were carried out on full FFPE sections of glass slides. Slides were stained using automated staining techniques, per manufacturer’s instructions (Ventana Medical Systems, Inc. Tucson, AZ), and were optimized and validated per Clinical Laboratory Improvement Amendments/College of American Pathologists (CAP) and International Organization for Standardization requirements. Staining was scored for intensity (0 = no staining; 1+ = weak staining; 2+ = moderate staining; 3+ = strong staining) and staining percentage (0%-100%). BC intrinsic subtypes were defined following American Society of Clinical Oncology/CAP guidelines.16 HER2-positive expression was defined as HER2 (4B5 clone) IHC 3+ staining or IHC 2+ with positive chromogenic in situ hybridization assay.

Real-world overall survival analysis
Caris CodeAI™ clinico-genomic database containing insurance claims data was used to calculate real-world overall survival (OS) from tissue collection to last contact. Kaplan–Meier curves were generated to calculate molecularly defined patient cohorts as previously described.13,14

Statistical analysis
Expression of examined genes were analyzed using Chi-square or Fisher Exact tests. Tumor microenvironment (TME) cell fractions were analyzed as described previously.13,14 A P value of < 0.05 was considered a statistically significant difference and P values were further corrected for multiple comparison using the Benjamini–Hochberg method to avoid type I error and an adjusted value (q value) of < 0.05 was considered a statistically significant difference. Spearman’s rank correlation was used to measure the association between variables where appropriate.

Results

Clinical and demographic patient information
A total of 19 697 BC samples with complete gene sequencing data available from Caris Life Sciences were identified (Figure 1). Of these 19 697 patient samples, 19 a459 (98.8%) were female with a median age of 61 years (range 20-89) and 238 (1.2%) were male with a median age of 67 years (range 33-89). The most common clinical subtype of MaBC was HR-positive/HER2-negative, representing 91 (38.2%) of the 238 MaBC samples (Figure 1A). HR-positive/HER2-negative was also the most prevalent subtype in FeBC (6097, 31.3%). Triple-negative BC (TNBC) was observed in 7.6% (18) of MaBC, as compared with 18.8% (3662) of FeBC, followed by HER2-positive [MaBC 6.3% (15) versus FeBC 6.9% (1344)]. The remaining cases of the 19 697 MaBC and FeBC samples with sequencing information did not have clinical subtyping data. Using self-reported race and ethnicity data, patients with FeBC were characterized as: white (69.8%), Black/African American (18.5%), Asian/Pacific Islander (3.5%), or other (4.5%); and not Hispanic or Latino (88.3%) or Hispanic or Latino (11.7%) (Figure 1B). A comparable, albeit less conclusive, trend was observed in MaBC, with much smaller sample sizes, particularly in the HER2-positive and TNBC group. Ductal BC was the most common histology across each molecular subtype for both MaBC and FeBC in all cases for which histology information was available. Full detail is provided in Figure 1B-C and Supplementary Table S1, available at https://doi.org/10.1016/j.esmoop.2026.106059.

Genomic mutational landscape across BC subtypes
The repertoire of mutations was displayed across the clinicopathological subtypes of BC, as shown in Figure 2. The mutational spectrum of HR-positive/HER2-negative BC (Figure 2A) revealed a lower incidence of mutations in MaBC compared with FeBC for TP53 (2.4% versus 31.2%), CDH1 (1.2% versus 17.5%), PIK3CA (30.2% versus 41.2%), ESR1 (5.8% versus 13.3%), PTEN (1.2% versus 6.6%), and an enrichment in mutations of DNA damage repair-associated BRCA2 (16.5% versus 4.2%) and in GATA3 (22.6% versus 14.1%). In HER2-positive BC (Figure 2B), there was a statistically significant difference in the rate of pathogenic mutations relevant to tumor suppression, which were more common in FeBC [TP53 (70.0% versus 26.7%)] over MaBC (Figure 2B). In TNBC, FeBC presented higher mutation rates compared with MaBC (Figure 2C) in TP53 (84.1% versus 38.9%), whereas MaBC had a significantly higher rate of mutation in CDH1 (16.7% versus 5.6%). Complete mutation data are presented in Supplementary Table S2, available at https://doi.org/10.1016/j.esmoop.2026.106059.

Immune profiling and tumor microenvironment analysis
The immune profiling analysis across all subtypes of BC showed no significant differences in the prevalence of PD-L1 IHC expression (22c3 and SP142 clones) and TMB-high (P = 0.52) (Figure 3A-C). Androgen receptor (AR)-IHC was the only marker observed to be significantly more prevalent in MaBC than FeBC across each molecular subtype (94.5% versus 83.6%: HR-positive/HER2-negative, 100% versus 80.3%: HER2-positive, 77.2% versus 23.2%: TNBC), but ARv7 splice variants between FeBC and MaBC were not dissimilar. Further characterization of the TME was carried out via analysis of inferred immune cell infiltrates by deconvolution of WTS and showed that MaBC had a relatively small but significant increased immune cell infiltration of M1 macrophages (Mϕ M1) in HR-positive/HER2-negative BC (2.7% versus 2.2%) and M2 macrophages (Mϕ M2) in TNBC (5.8% versus 3.0%) but significantly decreased—albeit relatively small—infiltration of dendritic cells in both HR-positive/HER2-negative BC and TNBC, (2.2% versus 2.6%, 1.6% versus 3.0%) compared with FeBC (Figure 4A, C). Immune-related gene expression in FeBC versus MaBC are shown in Figure 4D-F. In HR-positive/HER2-negative BC, PDCD1 (0.41 TPM versus 0.33 TPM), LAG3 (2.9 TPM versus 2.3 TPM), TNFSF14 (0.38 TPM versus 0.22 TPM), and CEACAM1 (24.5 TPM versus 17.9 TPM) were all significantly more prevalent in FeBC compared with MaBC (Figure 4D). In the HER2-positive cases, LAG3 in MaBC was the only immune-related gene with significantly higher prevalence than in FeBC (4.8 TPM versus 2.8 TPM) (Figure 4E). No significant differences were observed within the TNBC group (Figure 4F). Full data for each discussed figure are presented in Supplementary Tables S3, S4, and S5, available at https://doi.org/10.1016/j.esmoop.2026.106059.
Immune signatures IFNγ, T-cell-inflamed score, and MAPK activation score were all quantified for FeBC and MaBC across each molecular subtype (Figure 4G-I). Although there were some numeric differences between FeBC and MaBC, particularly for T-cell-inflamed score, there was no statistically significant observation for any immune signature tested. Additionally, gene expression profiles for both major histocompatibility complex (MHC) I and II were analyzed, but again, no significant differences were noted when comparing FeBC and MaBC across molecular subtypes. All data for immune signature quantitation and MHC class I and II gene expression are found in Supplementary Tables S6 and S7, available at https://doi.org/10.1016/j.esmoop.2026.106059.

CODEai™ real-world overall survival analysis
Real-world OS analysis was provided via Caris CODEai™ comparing BC subtypes within and across MaBC and FeBC patient cohorts for whom real-world OS data were available (Figure 5, Supplementary Figure S1, available at https://doi.org/10.1016/j.esmoop.2026.106059). In HR-positive/HER2-negative BC, FeBC had a median OS of 38.2 months [95% confidence interval (CI) 36.9-39.7 months] and MaBC had a median OS of 37.4 months (95% CI 31.2-61.7 months), P = 0.3 (Figure 5A). HER2-positive FeBC had a median OS of 43.1 months (95% CI 40.5-48.1 months) whereas HER2-positive MaBC had a median OS of 26.2 months (95% CI 10.4-54.1 months), P = 0.06 (Figure 5B). In TNBC, MaBC median OS was not reached (95% CI 8.8 months to not reached), and FeBC median OS was 21.2 months (95% CI 19.9-22.1 months), P = 0.1 (Figure 5C). Individual comparison between molecular subtypes for both MaBC and FeBC are shown in Supplementary Figure S1, available at https://doi.org/10.1016/j.esmoop.2026.106059. FeBC patients had significant differences in median OS between HR-positive/HER2-negative (38.3 months, 95% CI 36.9-39.7 months), HER2-positive (43.1 months, 95% CI 40.5-48.1 months), and TNBC (21.2 months, 95% CI 19.9-22.1 months), P = 0.0, Supplementary Figure S1A, available at https://doi.org/10.1016/j.esmoop.2026.106059. MaBC patients did not reach significance but had numerically trending median OS increase for HR-positive/HER2-negative BC (37.4 months, 95% CI 31.2-61.8 months) compared with HER2-positive BC (26.2 months, 95% CI 10.4-54.1 months, Supplementary Figure S1B, available at https://doi.org/10.1016/j.esmoop.2026.106059).

Discussion
In this study, we provide a comprehensive analysis of the genomic, transcriptomic, and immunologic landscape of MaBC compared with FeBC. In our series, MaBC accounted for ∼1% of all BC cases, consistent with its prevalence in the general population; the lower frequency of lobular histotype and CDH1 mutations in male patients, except in TNBC, along with the distribution of IHC-based subtypes, aligns with expected findings.17 Herein, we identify unique genetic alterations across BC subtypes in MaBC, with potential prognostic and therapeutic potential.
We evidenced an enrichment in HRD genes, in concordance with previously reported work.5 Although we could not confirm if HRD alterations are germline, they highlight a critical aspect of MaBC, suggesting tailored therapeutic implications—including the need to screen for germline HRD alterations and the potential treatment with poly (ADP-ribose) polymerase inhibitors (PARPi).18
HER2-positive MaBC shows higher rates of mutations in key tumor suppressor genes like BRCA1, BRCA2, PTEN, and GATA3, indicating distinct genetic vulnerabilities compared with FeBC. Although the mechanistic role of these alterations in determining trastuzumab resistance may be speculated, they may partly explain the observed clinical resistance of MaBC to HER2-directed treatments.19 Of note, a higher likelihood of trastuzumab resistance in MaBC in men undergoing neoadjuvant treatment has been reported: the rates of pathological complete response in one United States-based series was 16.1% in men versus 33.6% in women for HR-positive/HER2-positive BC P < 0.001); and 44% in men versus 53.2% in women for HR-negative/HER2-positive cancer (P = 0.42).20
The role of PIK3CA in determining resistance to HER2-directed antibodies has been largely documented, including in the advanced setting.21,22 In our series, 46.7% of MaBC and 33.3% of FeBC presented PIK3CA mutations, albeit the difference fell below the level of significance (P = 0.28). Similar findings have been reported recently by Kadamkulam Syriac et al.23
In HR-positive/HER2-negative FeBC, there was a higher prevalence of common mutations associated with endocrine resistance. The enrichment in ESR1 mutations in FeBC may suggest that the samples collected in MaBC had less pre-exposure to aromatase inhibitors, given that men with BC usually receive tamoxifen in the adjuvant setting, resulting in different rates of ESR1 mutations.17 However, such a difference might be associated with a separate tumorigenesis trajectory, driven by unique mechanisms that do not directly involve ESR1. For instance, HRD is a more common driver of tumorigenesis in MaBC, representing a separate mechanism that may disregard ESR1 mediated resistance in determining cancer progression. There is a higher prevalence of TP53 in FeBC, a cancer alteration associated with worse prognosis, portending primary endocrine resistance, and inferior outcomes in the advanced setting.24, 25, 26
Of interest, ESR1 and TP53 mutations appear mutually exclusive, but in MaBC they both represent a less relevant mechanism of endocrine resistance than in FeBC, where they contribute to the resistant phenotype in >50% of all cases.27 As such, alternative pathways are to be identified in MaBC, to identify and tackle endocrine resistance.27,28 Collectively, MaBC seems to progress through distinct trajectories of tumorigenesis and diverse endocrine resistance mechanisms. In MaBC, we also described an enrichment in mutations of BRCA2 (10.6% versus 4.4%), that has been proposed as a mechanism of resistance to endocrine and CDK4/6i therapy, potentially representing a more prevalent mechanism than in FeBC.29 These aggregated findings point to a broader disruption of regulated cell proliferation and DNA repair pathways across BC subtypes, with MaBC and FeBC exhibiting differential vulnerabilities that could inform therapeutic strategies.
The genetic portrait of TNBC suggests that its intrinsic heterogeneity may lead to multiple and unique resistance trajectories in MaBC.17 Given the limited data on triple-negative MaBC, the genomic differences are worthwhile for further exploration, specifically regarding the incidence of CDH1 mutation. Previous reports have implicated CDH1 mutation with the development of gastric cancer and lobular BC, with lifetime risks hovering between 40% and 80% for both men and women.30, 31, 32, 33 Importantly, lobular BCs in men are rare, representing 1.7% of all MaBC histological types.34
AR-IHC expression between MaBC and FeBC did not show significant differences among the entire patient population, regardless of molecular subtype. AR expression value carries significant clinical value as a pivotal marker in the pathogenesis of breast cancer.35 However, it should be noted that each subtype exhibited a trending increase in expression for MaBC. Previous studies have also reported high AR expression in MaBC,36,37 indicating this difference between FeBC may have significant prognostic and/or clinical importance. Given the role of AR in the biology of males, the numerically higher expression in males is expected.
The analysis of the immune markers and signatures used in clinical practice to infer immunogenicity did not reveal differences in the PD-L1 expression, TMB-high, IFNγ, T-cell inflammation, and MAPK scores. These findings are consistent with the literature, albeit some studies have reported a lower proportion of PD-L1 expression on immune cells in MaBC.38,39 However, a more granular assessment based on the transcriptomic patterns in our dataset suggests an increased immune cell infiltration of M1 Mϕ and M2 Mϕ with decreased dendritic cell content in MaBC, increased expression of MHC class II gene HLA-DQB2 and decreased expression of immune-related genes PDCD1, TNFSF14, and CEACAM1. The tumor environment appears immune-suppressive to immune-excluded, perhaps recapitulating the unique histology, subtype, and molecular landscape of MaBC. It is important to note, however, that although the difference in immune cell prevalence may be statistically significant, biological relevance would be better captured via evaluation of the co-localization patterns and functional assessment of the immune infiltrate of MaBC.
Caris CODEai™ OS analysis showed possible differences in the outcome of MaBC. Our findings are in line with previous studies, showing that men trend toward worse prognosis compared with FeBC.40 In our study, we speculate that the worse prognosis is related to resistance systemic treatments, specifically trastuzumab, albeit in a tiny subset of patients.34,41,42
Our study has limitations due to its retrospective design and small sample sizes of MaBC; samples were collected as clinically ordered, and OS analysis relied on insurance claims data. Not all patients in the CODEai™ analysis had molecular data, and some samples may originate from patients with atypical disease courses, a common issue in cancer genomic studies. In addition, the tumor subtype information is available as per the sample collection and we cannot capture whether tumor subtypes may have changed over time. Further, these samples were specifically referred for Caris molecular testing and were commonly sourced as advanced disease, possibly lending to selection bias. Comparisons were made within our dataset, requiring external validation. For both men and women, the time zero for all survival analyses was the date of tissue collection as a consistent biological event for all patients, which may or may not represent the date of recurrence; we expect any potential bias in the time difference from diagnosis to sample collection to apply similarly to both groups in our study. Selection of date of tissue collection as time zero was made to reduce inconsistencies across patients among other timepoints such as diagnosis date, which may depend on clinical interpretation, and treatment start, which depends on clinical decisions. A major strength of the study is being the first multi-omic survival analysis based on claim-based, well-annotated clinical data. Additionally, we provide data on HER2-positive and TNBC MaBC, areas with limited literature.
There are several areas to consider for future work to improve upon our characterization of MaBC. First and foremost, prospective analyses should be pursued for validation of the sex-specific genomic and immunologic patterns that have been described in this work as well as introducing clinical trials that explicitly stratify by sex and BC subtype. Additionally, we would encourage integration of germline alteration testing to fully describe distinct germline versus somatic HRD and BRCA2 aberrations as they pertain to PARPi strategies in MaBC. Finally, we would suggest dedicated characterization of the TME to describe the functional implications of our reported results, specifically regarding immunotherapy response in HER2-positive and TNBC MaBC.

Conclusion
In this study, we described the unique genomic, transcriptomic, immunological landscape, and survival of MaBC, highlighting potential areas of actionability and future new drug development. Continuous research efforts will be critical to confirm our findings and implement innovative strategies to understand, prevent, and treat the unique disease trajectory of MaBC.