This retrospective study was conducted as part of a broader research program aimed at characterizing the genetic landscape of breast cancer in Tunisian patients. Participants were randomly selected from the Department of Surgical Oncology at Mohamed Taher Maamouri University Hospital in Nabeul, Tunisia. Patients who underwent surgical resection between 2023 and 2024 were included if they had confirmed diagnosis of non-metastatic, histologically malignant epithelial breast tumor based on core needle biopsy. Written informed consent was obtained from all participants.

Among approximately 120 patients initially screened, twelve patients with confirmed TNBC were selected based on histopathological diagnosis. Immunohistochemistry analysis was performed using the fully automated Bond Max platform (Leica, Biosystems, Buccinasco, Milan, Italy). TNBC status was defined according to American Society of Clinical Oncology/College of American Pathologists (ASCO/CAP) guidelines as estrogen receptor (ER) expression <1% and progesterone receptor (PR) expression <1%. HER2 status was assessed as 0, 1+, or 2+ with negative chromogenic in situ hybridization (CISH) confirmation for equivocal cases. Clinical data were retrieved from patient medical records.

The study included eight surgically resected TNBC tissue samples collected after completion of neoadjuvant chemotherapy, along with matched adjacent non-tumor tissues. Four additional samples were obtained as formalin-fixed, paraffin-embedded (FFPE) specimens stored for less than 1 year. All Tissue sampling was performed under the supervision of a certified pathologist to ensure accurate tumor content assessment and proper handling. Immediately after surgical resection, all fresh tissue samples were immersed in RNAprotect® Tissue Reagent (QIAGEN, Hilden, Germany) and stored at −20 °C until nucleic acid extraction.

The study was conducted in accordance with the ethical principles of the Declaration of Helsinki and was approved by the Ethics Committee of Mohamed Taher Maamouri University Hospital in Nabeul (BC-01/2023).

Genomic DNA was extracted from fresh-frozen and FFPE tumor tissues using QIAGEN purification systems (QIAGEN, Hilden, Germany). DNA from fresh-frozen tissues was isolated using the QIAamp® DNA Mini Kit, while DNA extraction from FFPE samples was performed using the QIAamp® DNA FFPE Tissue Kit. All procedures were carried out strictly according to the manufacturer’s protocols to ensure high yield and purity.

DNA concentration was measured using the DeNovix QFX Fluorometer (DeNovix Inc., Wilmington, DE, USA) in combination with the Qubit™ dsDNA High-Sensitivity Assay Kit (Thermo Fisher Scientific, Waltham, MA, USA). DNA quality and integrity were evaluated using the Agilent 2100 Bioanalyzer (Agilent Technologies, Santa Clara, CA, USA), enabling assessment of fragment size distribution and sample suitability for downstream NGS applications.

Targeted NGS was carried out with the AmpliSeq for Illumina Cancer Hotspot Panel v2 (Illumina, San Diego, CA, USA), designed to amplify 207 specific amplicons encompassing roughly 2800 known mutations in the hotspot regions of 50 cancer-related genes. According to the manufacturer’s instructions, libraries were constructed from 100 ng of genomic DNA per sample using the AmpliSeq™ Library PLUS for Illumina® kit. Briefly, the procedure involved multiplex PCR amplification for target enrichment, followed by partial digestion of primer sequences and adapter ligation using the AmpliSeq™ CD Indexes Set A for Illumina® to assign unique barcodes to each library.

Libraries were purified using Agencourt® AMPure® XP beads (Beckman Coulter Inc., Brea, CA, USA). DNA concentration was quantified with the DeNovix QFX Fluorometer and fragment size profiles were further assessed using the Agilent 2100 Bioanalyzer (Agilent Technologies, Santa Clara, CA, USA). Libraries meeting quality requirements were normalized and pooled in equimolar ratios.

Prior to sequencing, the library pool was diluted to a final loading concentration of 8 pM and supplemented with a 5% PhiX Control v3 (Illumina) to enhance diversity and serve as an internal control. The pool was sequenced on an Illumina MiSeq platform equipped with a MiSeq Reagent Kit v2 (300-cycle) for 2 × 150 bp paired-end sequencing. The sequencing run was monitored using the Illumina Local Run Manager, with key metrics such as the percentage of bases ≥ Q30, mean coverage depth, and uniformity of coverage assessed against the manufacturer’s specifications. Only data passing these quality thresholds were subjected to subsequent bioinformatics analyses.

Raw sequencing data (FASTQ files) were processed using the DRAGEN Amplicon Pipeline on the Illumina BaseSpace Sequence Hub. Reads were aligned to the human reference genome (GRCh37/hg19). Somatic variant calling was performed using the DRAGEN Somatic Pipeline, which analyzes matched tumor-normal sample pairs to identify somatic mutations by filtering out germline variants present in the normal sample.

Functional annotation of the filtered variants was performed using SnpEff implemented on the Galaxy platform. The SnpEff workflow was configured with the hg19/GRCh37 reference database, enabling systematic classification of each variant according to its predicted molecular consequence. All shortlisted somatic variants were manually verified using the Integrative Genomics Viewer (IGV, version 2.16.2 07/2023) to inspect read alignment and rule out technical artefacts.

The final curated list of high-confidence variants was analyzed and visualized in RStudio (Version 2024.12.0+467). The maftools (version 2.12.0) and ComplexHeatmap packages (version 2.12.1) were used to generate comprehensive oncoplots, respectively, while additional custom plots were created using ggplot2 (version 3.5.1).

We constructed a protein-protein interaction (PPI) network to elucidate potential functional interactions among the mutated genes identified in our TNBC cohort. The network was generated using the STRING database (version 12.0) by inputting the respective gene symbols and executing the analysis with all default settings. Subsequently, the network was partitioned into clusters using the Markov Cluster Algorithm (MCL), applying the default inflation parameter of 3 to define protein complexes and functional modules. Finally, each cluster was annotated for biological pathway involvement via the built-in STRING enrichment analysis tool, with a focus on Gene Ontology (GO) pathways.

Descriptive statistical analyses were performed to summarize the demographic and clinical characteristics of the study cohort. Continuous variables were reported as mean ± standard deviation (SD). Categorical variables were expressed as frequencies and percentages. All statistical analyses were conducted using GraphPad Prism v10.0 (GraphPad Software, San Diego, CA, USA).

A total of twelve patients diagnosed with TNBC were included in this study. The demographic and clinicopathological characteristics are summarized in Table 1. The mean age at diagnosis was 58.3 ± 2.7 years, with ages ranging from 47 to 76 years. Tumor sizes varied between 4 mm and 40 mm, and the majority of tumors presented a high histological grade, with an SBR grade of 3 observed in most cases. Proliferative activity was markedly elevated, as reflected by a Ki-67 index ranging from 15% to 90%. Regarding treatment response, two patients achieved a pathological complete response (pCR = 100%), while four patients showed a good therapeutic response (>50%). The remaining patients exhibited partial response (<50%) or complete resistance (0%). Additionally, two patients developed tumor recurrence, one occurring 1 year after treatment and the other 5 years later. The cohort included eight fresh-frozen tissue samples and four FFPE samples, providing a representative set of routinely available clinical materials for molecular analysis.

Germline variant analysis across the twelve TNBC patients revealed a diverse set of inherited alterations, predominantly classified as benign or likely benign according to ClinVar, yet reflecting a broad genomic variability within the cohort (Table 2). The most frequently mutated gene was TP53 p.Pro72Arg (rs1042522) variant, detected in all patients (100%). Other genes with high mutation prevalence included CSF1R (rs2066933), FGFR3 (rs7688609), RET (rs1800861), KDR (rs7692791), and PDGFRA (rs1873778), each altered in over 80% of the cohort. Additional recurrently affected genes included APC (rs41115), FLT3 (rs2491231), EGFR (rs1050171), ERBB4 (rs839541), and PIK3CA (rs3729674), showing mutation frequencies ranging from 50% to 75%. Less frequently mutated genes, such as MET, ABL1, KIT, SMARCB1, ALK, ATM, FBXW7, VHL, HRAS, GNAS, IDH1, and NOTCH1, were identified in 8%–33% of patients (Figure 1a).

Analysis of variant types revealed that the majority of alterations were synonymous (42.6%, 84/197) or intronic (23.4%, 46/197), together accounting for approximately 67% of variants that are unlikely to affect protein structure Potentially consequential variants were less frequent: missense mutations represented 13.7% (27/197) of all germline variants, while putative loss-of-function alterations, including frameshift insertions and deletions, were rare (1.0% combined). Additionally, 7% of variants affected 3′UTR regions, which can influence gene transcription and post-transcriptional regulation. Collectively, around 23% of the germline variants identified may play a pivotal role in TNBC susceptibility (Figure 1b).

This profile highlights a recurrent oncogenic germline variant landscape within our TNBC cohort, underscoring the need for investigation in larger patient populations to better elucidate their contribution to inherited susceptibility in this aggressive breast cancer subtype.

Furthermore, the analysis of single nucleotide variation (SNV) substitution types revealed that the most frequent substitution observed was G>A, which accounted for 28% of all SNVs (n = 51). This was followed closely by the A>G substitution, which represented 19.8% of the total (n = 36), and C>T substitutions, contributing 14.3% (n = 26). Cumulatively, these three transition types (G>A, A>G, and C>T) constituted 62.1% of the total SNV burden. Conversely, transversions (purine to pyrimidine or pyrimidine-to-purine changes) were less frequent. The intermediate substitutions included T>C (12.1%, n = 22), G>C (7.1%, n = 13), and T>G (6%, n = 11). The least common substitution type was A>C, observed only once (0.5%), followed by other low-frequency events such as T>A (2.2%, n = 4) and C>A (2.2%, n = 4). The observed signature, characterized by a high proportion of G>A and A>G transitions, suggests the presence of a specific underlying mutational process shaping the germline variant landscape in the analyzed TNBC samples (Figure 1c).

Somatic variant analysis in the twelve TNBC patients revealed a heterogeneous mutational landscape affecting multiple cancer-related genes (Table 3). A total of 32 deleterious somatic variants across 20 oncogenes were identified. Analysis of the somatic mutational landscape revealed a median of 3 deleterious mutations per tumor (Figure 2a). The vast majority of variants were single nucleotide polymorphisms (SNPs), with a strong predominance of C>T transitions, followed by T>C and C>G substitutions, consistent with mutational signatures commonly observed in solid tumors. Insertions (INS) and deletions (DEL) collectively accounted for a minor fraction of the total variants. These somatic variants exhibited variable predicted pathogenicity according to ClinVar and AMP classification, reflecting the complex genomic heterogeneity characteristic of TNBC (Figure 2c).

The most frequently altered gene was TP53, with pathogenic or likely pathogenic missense variants observed in 58% of patients, predominantly affecting exons 5–8, which encode the DNA-binding domain (DBD). Notable recurrent TP53 variants within the DBD included p.Val173Leu, p.Val157Phe, p.Ser127Phe, p.Arg273Cys, p.Tyr220Cys, p.His193Tyr, p.Asn239Ser, as well as a novel splice region deletion (c.560-6_560-1del) identified in our cohort (Figure 3). Other frequently mutated tumor suppressors and oncogenes included APC, STK11, SMAD4, RET, CDKN2A, MET, and VHL, with mutation frequencies ranging from 8% to 33%, highlighting recurrent alterations in key cancer-related pathways (Figure 2b). This profile underscores the concerted disruption of central tumor suppressor and oncogenic pathways, including p53 signaling, Wnt/β-catenin, and cell cycle regulation, in this TNBC cohort.

Variant type analysis revealed that missense mutations were the most common somatic alteration, followed by frameshift insertions/deletions, nonsense mutations, in-frame indels, and splice site alterations (Figure 2c). Tier 1 (pathogenic or likely pathogenic) mutations were identified in key oncogenes and tumor suppressors, including TP53, STK11, RET, and VHL, highlighting potential driver events that may contribute to TNBC tumorigenesis and progression (Figure 2b).

Correlations between the mutational profile and treatment response were observed. Several genes harbored multiple independent somatic variants in patients with differing therapeutic outcomes. Variants in APC, SMAD4, RET, and ATM were detected in patients who achieved a partial response to treatment (pCR >50%) but did not reach a complete response. In contrast, multiple missense variants in TP53 and single missense variants in CDH1, FGFR2, and KIT were predominantly found in patients with poor treatment response (pCR <50%). Notably, an in-frame deletion in NOTCH1 (p.Val1578del) was exclusively detected in two patients who later developed disease recurrence, one at 1 year and the other at 5 years, indicating a possible link between this variant and long-term relapse risk. Furthermore, a missense mutation in AKT1 (p.Glu17Lys, exon 4) co-occurred with the TP53 pathogenic variant p.Val173Leu in a treatment-resistant tumor (pCR = 0), highlighting a potential cooperative effect involving AKT1–TP53 axis dysfunction in mediating complete resistance (Figure 2b). These findings indicate that the distribution and type of somatic variants may correlate with treatment response in TNBC, highlighting candidate genes that could influence therapeutic outcomes and potentially serve as predictive biomarkers for relapse and resistance.

To elucidate the functional relationships among altered genes in our TNBC cohort, we constructed separate PPI networks for germline and somatic mutations using the STRING database.

The germline network, built from recurrently mutated genes (frequency >50%), comprised 12 nodes and 34 edges, significantly more than the 13 expected by chance (PPI enrichment p-value <0.001). This network was highly interconnected, with an average node degree of 5.67 and a clustering coefficient of 0.817, and featured TP53 as a central hub (Figure 4a).

The somatic network, incorporating all deleterious mutations, demonstrated even greater connectivity. Its 19 nodes and 102 edges (expected: 27; p < 0.001) formed a dense architecture, evidenced by an average node degree of 10.7 and a clustering coefficient of 0.799. TP53 again emerged as a central hub, underscoring its pivotal role in TNBC tumorigenesis across both genetic layers (Figure 4b).

Cluster analysis using the MCL identified a major functional module in both germline and somatic networks, significantly enriched for pathways including central carbon metabolism in cancer and transmembrane receptor protein tyrosine kinase activity, indicating coordinated involvement of these altered genes in metabolic regulation and oncogenic signalling pathways.

Gene Ontology (GO) enrichment analysis of germline-mutated genes revealed a highly significant overrepresentation of pathways associated with growth factor signalling and kinase-mediated regulatory mechanisms (Figure 4c). The top enriched Biological Process terms (False Discovery Rate, FDR <1.0 × 10⁻⁹) included transmembrane receptor protein tyrosine kinase signalling pathway, positive regulation of cell proliferation, and regulation of MAPK cascade. A considerable proportion of germline variants mapped to these signalling-related categories, suggesting an inherent predisposition to perturb key regulatory pathways governing cellular growth, survival, and proliferation. These findings indicate that inherited alterations may prime oncogenic signalling networks, thereby creating a favourable molecular context for TNBC initiation or early tumour development.

In contrast, GO enrichment analysis of somatically mutated genes revealed a strong overrepresentation of pathways involved in cell cycle control and maintenance of genomic integrity (Figure 4d). The most significantly enriched Biological Process terms (FDR <1.0 × 10⁻⁹) included mitotic cell cycle process, DNA repair, and response to DNA damage stimulus. Notably, a high number of somatically altered genes contributed to these key pathways (ranging from 9 to 12 genes per term), reflecting a concerted disruption of mechanisms essential for accurate cell division and genome stability. This pattern underscores that in TNBC, somatic mutations predominantly target processes driving uncontrolled proliferation and defective DNA damage responses, thereby promoting tumour progression.

Genetic analysis of germline variants in breast cancer, including TNBC, has traditionally focused on pathogenic variants in BRCA1/2, as these genes are well-established drivers of hereditary breast cancer. However, other oncogenes and tumor-suppressor genes remain comparatively underexplored, despite growing evidence suggesting their important contribution to breast carcinogenesis [14]. In this study, we identified a spectrum of recurrent germline variants across several cancer-associated genes in patients with TNBC, including TP53, CSF1R, FGFR3, RET, KDR, PDGFRA, APC, FLT3, EGFR, ERBB4, PIK3CA, MET, and ABL1.

In the absence of a matched control group, distinguishing TNBC-specific germline variants from population-level polymorphisms remains challenging. To partially address this limitation, we compared variant frequencies observed in our cohort with allele frequencies reported in the Genome Aggregation Database (gnomAD). Several variants identified in our cohort are also present at high frequencies in the general population, including FGFR3 (rs7688609, 100% vs. 99%), PDGFRA (rs1873778, 92% vs. 99%), CSF1R (rs2066933, 100% vs. 77%), and FLT3 (rs2491231, 83% vs. 74%). Similarly, relatively high frequencies were observed for RET (rs1800861, 92% vs. 76%), TP53 (rs1042522, 100% vs. 70%), APC (rs41115, 75% vs. 62%), and EGFR (rs1050171, 75% vs. 56%) compared with gnomAD. These observations suggest that many of these variants likely represent common germline polymorphisms rather than TNBC-specific susceptibility variants.

In contrast, other variants identified in our cohort appear at markedly lower frequencies in gnomAD, including KDR (rs7692791, 83% vs. 54%), ERBB4 (rs839541, 58% vs. 26%), PIK3CA (rs2230461, 33% vs. 6%), STK11 (rs2075606, 50% vs. 25%), MET (rs35775721, 42% vs. 5%), and ABL1 (rs754813636, 25% vs. <0.001). Although these variants cannot be definitively associated with TNBC susceptibility without a matched control group, their relatively lower population frequencies may suggest a potential contribution to disease risk or tumor biology in this population.

Given the limited availability of genomic data for North African populations, we compared our findings with those recently reported by our group in a hepatocellular carcinoma (HCC) cohort analyzed using the same sequencing panel and analytical pipeline [15]. Several germline alterations were detected in both studies, although their frequencies differed between the two cancer types. These observations may indicate that these variants represent common polymorphisms within this population, or alternatively, that their combined presence reflects a genetic background that could influence cancer susceptibility.

Notably, several of these germline variants identified in both cancer types—including FGFR3 (rs7688609, COSM4533173), PDGFRA (rs1873778, COSM7410554), RET (rs1800861, COSM4418405), TP53 (rs1042522, COSM250061), APC (rs41115, COSM3760869), EGFR (rs1050171, COSM1451600), KDR (rs1870377, COSM149673), ERBB4 (rs839541, COSM19690034), PIK3CA (rs2230461, COSM328028), STK11 (rs2075606, COSM6666958), and MET (rs35775721, COSM1579024) — have already been reported as somatic mutations in the COSMIC database. Although several of these variants are classified as synonymous substitutions and are generally considered likely benign, they have nevertheless been reported to be implicated in key molecular pathways involved in the development of multiple cancer types. Examples include FGFR3 (rs7688609) in glioblastoma [16], RET (rs1800861) in thyroid carcinoma [17], KDR (rs7692791) in HCC [18], APC (rs41115) in adenomatous polyposis [19], and FLT3 (rs2491231) in TNBC [20].

The most frequently detected germline variant in our cohort was the well-characterized missense variant Pro72Arg (rs1042522, COSM250061) located in exon 4 of the tumor suppressor gene TP53. This variant, which results in a proline-to-arginine substitution at codon 72, is among the most widely distributed TP53 variants and has been reported across numerous cancer types, including breast, lung, colorectal, ovarian, and hepatocellular carcinoma in several ethnic [21]. Functionally, two isoforms exhibit distinct biological properties. The Arg72 (R72) variant has been shown to induce apoptosis, as well as influence cell migration, invasion, and metastatic potential more effectively than the Pro72 (P72) variant. Conversely, the P72 isoform is associated with enhanced DNA-repair capacity and increased cellular survival under genotoxic stress [22, 23]. In breast cancer, the presence of mutant p53 carrying the R72 variant has been significantly associated with poorer clinical outcomes [24]. Furthermore, recent case–control studies conducted in the Tunisian population have reported associations between this polymorphism and increased risk to chronic lymphocytic leukemia and cervical cancer [25, 26]. These findings indicate that, while not classified as pathogenic, Pro72Arg warrants further investigation to clarify its potential modulatory role in cancer susceptibility and tumor behaviour.

Additionally, we identified two recurrent 3′UTR variants in CSF1R gene (rs2066934 and rs2066933) in all samples of our cohort. Variants in the 3′UTR can alter post-transcriptional regulation by creating or disrupting microRNA recognition elements (MREs), thereby modulating mRNA stability and translation; In silico analyses have specifically predicted that rs2066934 may affect binding of numerous miRNAs and thus has the potential to substantially alter CSF1R expression [27]. Notably, growing evidence from high-throughput cancer sequencing studies indicates that these two CSF1R polymorphisms (rs2066934 and rs2066933) occur at relatively high frequencies, highlighting their emerging relevance in cancer genomics [28–30].

Taken together, the recurrent detection of these hotspot polymorphisms across different cancers highlights the need for well-designed case-control studies and underscores the importance of investigating potential gene-gene interactions to better clarify the contribution of these variants to cancer susceptibility.

Furthermore, PPI enrichment analysis revealed that the set of germline-altered genes identified in our study is significantly involved in major oncogenic signaling pathways. Notably, these genes clustered within the central carbon metabolism in cancer and PI3K/AKT signaling pathways, both crucial for metabolic reprogramming and survival in cancer cells [31]. Metabolic reprogramming, including enhanced glycolysis and glutaminolysis, is now considered a hallmark of cancer and supports anabolic processes and adaptation to stress [32]. In addition, enriched biological processes in our data set included protein autophosphorylation, transmembrane receptor protein tyrosine kinase signalling, and the VEGF signalling pathway, which regulate activation of receptor tyrosine kinases mechanisms frequently implicated in tumor progression and angiogenesis [33]. The involvement of these processes suggests that germline variants may modulate not only oncogenic signalling but also angiogenic responses and cellular proliferation. Collectively, these subtle germline alterations, despite being often labelled as benign, might perturb critical signalling networks and contribute to cancer susceptibility by priming cells for enhanced responsiveness to oncogenic stimuli.

In addition to germline alterations, we investigated the spectrum of somatic mutations and their potential association with clinical features, particularly response to neoadjuvant therapy. As expected, TP53 emerged as the most frequently mutated gene, with nine deleterious alterations detected in 58% of tumor samples, consistent with its well-established role as the dominant driver of TNBC [34]. These missense mutations, all located within the DNA-binding domain (DBD), impair the oncosuppressive function of p53. These alterations were classified into two major categories with distinct functional and clinical implications. Mutations affecting residues directly involved in DNA interaction were classified as ‘contact mutants’. In contrast, substitutions such as V173L, R175H, and Y220C represent ‘structural mutants’, which destabilize the core DBD architecture, leading to its misfolding under physiological conditions [35, 36]. Critically, this structural misfolding is strongly associated with dominant-negative effects over p63 and p73 function and gain-of-function activities, which together confer enhanced resistance to chemotherapy-induced apoptosis compared to contact mutants [37, 38].

Our findings align with this functional dichotomy. Contact mutations such as p.Val157Phe, p.Arg273Cys, and p.Asn235Asp were detected in patients who achieved a pathologic complete response, while p.Asn239Ser was observed in a patient with a good therapeutic response. Moreover, multiple deleterious TP53 alterations, including p.His193Tyr, p.Ser127Phe, p.Val157Phe, and a novel splice-region deletion (c.560-6_560-1del), were only detected in patients with partial responses. Although these observations cannot establish a causal relationship, they may reflect cumulative structural destabilization and dominant-negative effects that impede chemotherapy efficacy.

In contrast to the favourable responses observed with contact mutations, structural TP53 mutants showed a markedly different pattern in our cohort. Notably, the p.Tyr220Cys (Y220C) substitution, one of the most destabilizing structural variants of the DBD, was identified in a patient who experienced tumor recurrence 1 year after treatment, consistent with the strong association of Y220C with impaired apoptotic signalling and poor clinical outcome [39]. Similarly, the p.Val173Leu (V173L, rs876660754) variant, another well-characterized structural mutant, was detected in a patient who exhibited complete resistance to neoadjuvant chemotherapy [36]. Moreover, previous studies have reported that the V173L variant co-occurred with an activating AKT1 (E17K, rs121434592) alteration, which may further promote pro-survival signalling and contribute to the observed chemo-resistance [40]. The AKT1 E17K mutation (rs121434592) has been widely reported to play an important role in tumor development and chemotherapy resistance across solid tumor types, including breast cancer [40, 41].

Furthermore, we identified the same NOTCH1 in-frame deletion (p.Val1578del) in both patients who relapsed. This variant has been reported to affect part of the PEST domain and is predicted to impair degradation of the NOTCH intracellular domain (NICD), leading to its prolonged nuclear persistence and sustained oncogenic signalling [42]. Notably, the patient who relapsed after 1 year also carried the TP53 Y220C structural mutant, suggesting a possible cooperative effect between dysfunctional p53 and persistent NOTCH1 activation that may contribute to early chemo-resistant relapse. In contrast, we hypothesize that the late recurrence at 5 years was driven solely by the NOTCH1 PEST-domain deletion, raising the possibility that NOTCH1 activation alone may sustain minimal residual disease and fuel long-term relapse. Nevertheless, they are consistent with previous reports indicating that TNBC harboring PEST-domain NOTCH1 mutations are sensitive to γ-secretase inhibitors [43]. Given the limited sample size, these observations should be interpreted with caution. Furthermore, comparison of the somatic mutations identified in this study with those reported in our previous work on HCC [15], reveals distinct mutational landscapes across the two tumor types, underscoring the context-specific nature of oncogenic alterations and highlighting the importance of cancer-tailored genomic screening strategies.

This study has some limitations that should be acknowledged. First, our findings are derived from a single-institution cohort with a relatively small sample size, reflecting the low prevalence of TNBC (10%–15%) among breast cancer cases in our population. While this provides valuable preliminary insights, it may limit the generalizability of the results and introduce potential selection bias. Future studies involving larger, multi-center cohorts are required to validate and extend these findings.

Second, the use of a targeted gene panel, while clinically focused, and cost-effective, restricts the discovery of novel alterations outside the predefined regions. Notably, this panel does not cover key TNBC-associated genes such as BRCA1/2, which are critical for a complete characterization of TNBC mutational profiles in Tunisian cohort. Broader genomic approaches, such as whole-exome or whole-genome sequencing, would provide a more comprehensive understanding of the genetic determinants of TNBC and enable functional characterization of newly identified variants. Expanding sequencing efforts to larger cohorts and additional genomic regions and other breast cancer subtypes, will be essential for uncovering population-specific mutations and improving clinical management in Tunisian breast cancer patients.

TNBC is a highly aggressive breast cancer subtype with limited targeted therapeutic options.

First exploratory profiling of germline and somatic variants in Tunisian TNBC patients.