Keratin as a Potential Prognostic Biomarker for Breast Cancer Progression: An Evidence-Synthesizing Narrative Review.

Introduction
BC manifests when a specific cell located in either ducts or lobules proliferates uncontrollably.1 This condition is of particular concern because female sex is considered one of the primary risk factors for developing BC, which has emerged as a leading cause of mortality worldwide.2 Recent global cancer statistics demonstrate that approximately 2,261,419 new cases of BC were diagnosed in 2020, with 685,000 deaths occurring worldwide in the same year.3
During tumor progression, malignant transformation induces a cascade of structural rearrangements, primarily driven by the EMT process. This process enables cells to acquire migratory and invasive properties that promote metastasis, a major cause of BC-related mortality. The EMT process involves cytoskeletal remodeling, enzyme secretion for extracellular matrix degradation, and altered microRNA expression, all of which facilitate cancer cell dissemination and therapy resistance.4
The prevailing diagnostic tools employed in the contemporary context to identify BC include imaging tests (eg, mammograms, ultrasounds, and magnetic resonance imaging) to examine breast tissue for any unusual findings. Markers such as hormone receptors, including estrogen receptor (ER), progesterone receptor (PR), and human epidermal growth factor receptor 2 (HER2), are found at the site where the tissue sample is extracted.5 However, the application of hormone markers has been demonstrated to be ineffective as hormone markers. First, they are not present in all BC subtypes, such as TNBC. TNBC is characterized by the absence of ER and PR expression, as well as HER2 amplification, resulting in a highly heterogeneous disease with limited well-defined molecular targets.6 Second, it is also possible that the expression of hormone receptors may change during disease progression. The study revealed a statistically significant change in 32.4% of patients in the ER group (P < 0.001), 40.7% of patients in the PR group (P < 0.001), and 14.5% of patients in the HER2 group (P = 0.44). Furthermore, females with ER+ primary tumors that changed to ER- tumors presented a substantial 48% elevated risk of mortality.7 Third, the presence of hormone receptors does not serve as an indicator of therapy resistance. Studies have shown that ER+ status is associated with a reduction in the efficacy of hormone therapy.8
Given the limitations of traditional hormonal biomarkers, recent attention has shifted toward cytoskeletal components specifically, intermediate filaments (IFs) such as vimentin, nestin, and keratin as alternative biomarkers. IFs maintain cell integrity and structure but are also dynamically remodeled during EMT, linking them directly to cancer aggressiveness, invasion, and therapeutic response. Alterations in the expression of these filaments have been correlated with BC progression, metastasis, and chemoresistance.4
Among cytoskeletal biomarkers, keratins are especially promising for clinical translation as they are lineage-specific, structurally stable, and routinely assessed in diagnostic pathology using immunohistochemistry (IHC). Different from hormone receptors, keratin expression depicts epithelial differentiation states and cytoskeletal remodeling that persist during disease progression, metastatic dissemination, and exposure to treatment. These features position keratins at the interface between tumor biology and clinical decision-making.
Here the role of keratin as a biomarker in BC was described in-depth, with significant interest in its prognostic and predictive effects in various molecular subtypes. The selected keratin subtypes addressed in this review (K5/6, K14, K17, K19, and K8/18) are considered the predominant basal and luminal keratin groups in BC and are the most clinically significant, with the strongest supporting evidence in the literature. The accumulating clinical and experimental data suggest that certain keratin subtypes are consistently associated with tumor aggressiveness, therapeutic response, and clinical outcomes that warrant their consideration beyond descriptive epithelial markers. Although the data for certain subtypes of keratin in predicting therapeutic response (predictive biomarkers) are limited, the growing evidence indicates that their consistent association with disease aggressiveness and clinical outcomes supports their broader role as prognostic markers. Current oncology frameworks emphasize that biomarker-guided therapy is best accomplished when a biomarker robustly connects tumor biology to actionable clinical decisions and not as a simple molecular feature only.
However, only a limited number of studies summarizing the use of potential keratin, both in the form of individuals and/or panels, as diagnostic and prognostic biomarkers for BC have been published. This review aims to address the need for clinical application to determine which keratin is the most relevant and accurate prognostic tool across various BC subtypes, contributing to a deeper understanding of the role of keratin as a biomarker for treatment approaches.

Materials and Methods

Eligibility Criteria
This review was conducted as a narrative synthesis and full PRISMA guidelines were not applied. Instead, selected elements of the PRISMA flow diagram were adapted to enhance transparency in the search, screening, and study selection processes. The literature search was performed in two screening phases. The first phase entailed abstract/keyword screening, whereas the second phase involved full-text screening. In the first phase, studies from PubMed were selected using medical subject heading (MeSH) terms and limited by date and text availability (Table 1). Additional studies were sourced from Scopus and filtered by year, subject, document type, language, and open access.

Inclusion and Exclusion Criteria
The PICO framework for keyword selection in this study is structured as follows: population (P) focuses on the disease population and setting (DPS), with a specific focus on BC patients of all types within a global setting. The exposure (E) under consideration is keratin (or intermediate filaments). For comparison (C), the study evaluated various prognostic tools used for BC assessment. The outcome (O) aims to determine the expression of keratin and its prognostic value in BC patients, while the timeframe (T) for the study is set at 10 years, ensuring relevance to past trends and data. Finally, the design (D) of the research follows either a cross-sectional or human study approach, providing comprehensive insights into the relationship between keratin expression and BC prognosis.

Data Source and Search Strategy
In the review, Rayyan.ai was used to evaluate the eligibility of studies and to identify any potential duplications or irrelevances. Rayyan.ai categorizes articles as “include”, “exclude”, or “maybe” based on eligibility criteria. The inclusion criteria of this study are papers that are the results of original articles (human, in vivo, in vitro), titles, keywords and abstracts of sources that are consistent with the objectives of the study. In addition to the abstract, we included only journals that provided information about the sample analyzed, the investigated biomarkers, and the comparisons between groups shown in Figure 1 about identification of studies via databases and registers.

Results
The studies (Tables 2–4) reviewed a total of 79 studies, which were categorized based on their design into three groups: in vitro, in vivo, and human studies. Each study addressed distinct scientific objectives. Among these studies, eleven in vitro investigations focused on elucidating keratin-associated signaling pathways, providing insights into how keratins drive cancer progression at the cellular level, and laying the foundation for clinical applications. Furthermore, six in vivo studies examined therapeutic interventions, utilizing preclinical models to test treatment strategies targeting keratin-expressing tumors. These studies included the evaluation of therapeutic agents delivered through keratin-targeting systems, which offers valuable guidance for future treatment development. In addition, 62 human studies investigated the prognostic significance of keratin subtypes in human BC. These studies analyzed keratin expression patterns to classify subtypes and predict clinical outcomes. Collectively, these studies contribute to a more comprehensive understanding of keratin function in cancer and its potential as a therapeutic and prognostic marker.

McGinn et al (2020) demonstrated that the upregulation of K5 activates the Wnt/β-catenin pathway, disrupts cell adhesion, and enhances endocytosis in the luminal (ER+), basal-like triple-negative and HER2+ BC subtypes. K5 stabilizes β-catenin through these pathways, enabling its nuclear translocation and TCF/LEF-mediated Wnt1 activation. This process sustains cancer stem cell (CSC) properties. K5/β-catenin interactions weaken E-cadherin, thereby reducing cell–cell adhesion and promoting invasion. Enhanced endocytosis and trafficking further alter signaling dynamics, thereby driving tumor progression. Taken together, these mechanisms increase invasion, CSC enrichment and EMT, and predicting a poor prognosis in ER+ and TNBC patients.9
Wang et al (2020) reported that, in TNBC and luminal basal-like subtypes, the simultaneous upregulation of K5, K6, K14, and K17 under anilin expression triggers pathways involving the regulation of nuclear actin, the induction of E-cadherin, and the maintenance of CSCs. Anillin modulates RNA polymerase II activity, affecting stemness genes, while its loss upregulates E-cadherin, reducing migration, and invasion. These pathways promote migration, invasion, metastasis, and cytoskeletal remodeling by sustaining CD44 (high)/CD24 (low) populations and mammosphere formation through TBX18 and SOX9 downregulation. High anillin levels are correlated with poor survival in TNBC patients.10
Verma et al (2022) and Padmanaban et al (2020) demonstrated that, in basal-like TNBC, high K14 expression activates H3K27me3-mediated transcriptional activation and Rho/ROCK/integrin signaling, thereby driving cytoskeletal reorganization and invasion. Specifically, H3K27me3 blocks the SPI repressor and allows RNA polymerase II recruitment, promoting KRT14 expression. This activation enhances migration, invasion, and metastasis, establishing K14 as a robust prognostic biomarker that predicts poor prognosis.11,12
Izdebska et al (2021) and Lin et al (2020) demonstrated that, in TNBC, luminal A (ER+) and luminal B (ER−/HER2+) cancers, the upregulation of K19 activates the HER2/ERK, Wnt/β-catenin, EMT, and ER stress signaling pathways. Matrix metalloproteinase (MMP)-9 degrades extracellular matrix (ECM) proteins, creating paths for metastasis, while the upregulation of vimentin and N-cadherin enhances EMT. K19 also stabilizes HER2/ERK and Wnt signaling, disrupts E-cadherin and modulates the unfolded protein response. Collectively, these processes promote migration, invasion, EMT, and cytoskeletal remodeling, all of which predict poor outcomes when MMP-9 levels are high.13,14
Zhao et al (2023) reported that in luminal TNBC, the downregulation of KRT7-AS and the subsequent activation of KRT7 stimulate the PTEN/PI3K/AKT pathway and degrade KRT7/FOXA1. Typically, KRT7-AS stabilizes PTEN, thereby suppressing AKT/NF-κB-driven proliferation and oncogenic FOXA1 expression. However, its loss enhances migration, invasion, metastasis, immune evasion, and chemoresistance, predicting a poor prognosis.15
Butler et al (2020) revealed that, in TNBC, KRT18 is silenced via DNMT3A-mediated promoter hypermethylation, which activates the Wnt/β-catenin and EMT pathways. Loss of KRT18 increases N-cadherin and reduces epithelial integrity, thereby increasing migration, invasion, and metastasis. Azacytidine counteracts this by inhibiting Wnt-3/4, GSK-3, and β-catenin, thereby reducing stemness and invasion. It also decreases VEGF secretion, which impairs tumor vascularization. Overall, DNMT3A-induced KRT18 silencing promotes EMT and an aggressive phenotype is associated with a poor prognosis.16
Saha et al (2018, 2019) demonstrated that KRT19 downregulation in TNBC activated the Src/GSK3β, Wnt/β-catenin/Notch, and HER2/ERK/SP1 pathways. The KRT19/β-catenin/RAC1 complex suppresses NUMB, thereby enabling Notch activation, enhancing stemness, and promoting invasion. KRT19 also stabilizes HER2 to promote proliferation and inhibits Src/GSK3β phosphorylation, thereby removing the brakes on stemness. These interactions increase migration, invasion, proliferation, stemness, EMT, and metastasis, predicting a worse prognosis.17,18
Yan et al (2024) demonstrated that, in TNBC, KRT81 upregulation activated focal adhesion signaling (FAK/Paxillin) and immune- and metabolic-related pathways, such as oxidative phosphorylation and glycolysis. These findings support migration, invasion, proliferation, and immune evasion, with elevated KRT81 predicting a poor prognosis.19
Key findings, including those of Saghaeidehkordi et al (2021), demonstrated that the upregulation of K1 enables a targeted drug delivery pathway in TNBC via NOD/SCID mice. A PDC binds to K1, which is expressed on TNBC cells and enters them via receptor-mediated endocytosis. This releases doxorubicin specifically in the acidic tumor environment. This pathway enhances tumor-specific drug delivery, reduces doxorubicin accumulation in healthy tissues, improves tumor suppression, and minimizes side effects. Consequently, high K1 expression in TNBC indicates a poor prognosis.However, the use of PDC improves therapeutic precision and could be adapted for other chemotherapies.20
Machado et al (2020) reported that in dogs with TNBL mammary tumors, high expression of K5/6 is associated with a signaling axis in which loss of ERβ promotes basal-like differentiation by driving EMT through loss of E-cadherin and upregulation of vimentin. K5/6 marks basal/myoepithelial cell layers and is associated with more aggressive tumor behavior, higher mitotic rates, and lymph node metastasis. Consequently, K5/6+ basal-like tumors predict a poor prognosis because of their rapid progression and metastasis. In contrast, triple-negative nonbasal-like tumors (K5/6-) are aggressive but less well characterized.21
Eivani et al (2016) reported on canine mammary carcinomas and examined the prognostic relevance of K5/6 and K7. While the basal marker K5/6 typically indicates aggressive, hormone receptor-negative tumors and poor outcomes in humans, this study revealed that neither K5/6 nor K7 had significant prognostic value in dogs. Furthermore, no clear link with EMT was observed, indicating species-specific differences in keratin-related tumor behavior.22
Bado et al (2017) demonstrated that, in mice with K14CreERβp53 tumors (BLBC), high K14 expression emerges when ERβ is lost and p53 is knocked down. The loss of ERβ disrupts epithelial maintenance, shifting differentiation toward basal-like clones and promoting EMT, as evidenced by decreased E-cadherin expression and increased vimentin, N-cadherin, and α-SMA expression. This shift enhances invasion and metastatic potential, indicating a poor prognosis when ERβ levels are low and K14 levels are high. Conversely, maintaining ERβ alongside lower levels of K14 favors epithelial features and better outcomes.23
WalyEldeen et al (2024) reported that the interplay between K14, K18, and FBLN2 determines tumor behavior across claudin-low, luminal A, and basal/HER2+ BC subtypes in C57BL/6 mice and felines. FBLN2 interacts closely with KRT14 and integrins (ITGβ1/ITGα3) to maintain the basement membrane (BM) and suppress invasive potential by stabilizing the ECM. High FBLN2 expression upregulates KRT14 and downregulates KRT18, preserving a basal phenotype with a favorable prognosis in the early stages of the disease. In contrast, low FBLN2 expression shifts cells toward a luminal phenotype (KRT18+) with BM loss, promoting invasiveness and a poor prognosis in advanced stages.24
Dagher et al (2020) revealed that, in feline triple-negative basal-like carcinomas, high K14 expression is associated with increased infiltration of regulatory T cells (Tregs) marked by FoxP3, creating an immunosuppressed microenvironment. This suppresses antitumor immunity, facilitating tumor progression, escape, growth, lymph node metastasis, and advanced stages. Therefore, K14+ Treg-rich tumors, which are characteristic of the BLIS-like subtype, predict the worst outcomes, whereas luminal tumors with lower Treg infiltration are associated with longer survival.25
Engström et al (2017) demonstrated that, in luminal A/B and TNBC, the upregulation of K5 and K14 helps to maintain stem-like properties while also altering cell adhesion and migration. These pathways promote tumor invasion by driving cytoskeletal remodeling and EMT. The study also revealed that the status of the BM can be used to categorize luminal HER2- tumors, with BM+ luminal HER2- patients having a better prognosis than BM-patients do.26
Abdelrahman et al (2017) demonstrated that, in TNBC, high K5/6 expression stabilizes the cytoskeleton and facilitates metastasis through EGFR/PI3K-driven proliferation. K5/6 act as markers of high-risk TNBC subsets, promoting EMT via cytoskeletal remodeling and loss of epithelial markers, which correlates with aggressive clinical behavior. Thus, K5/6 upregulation predicts a poor prognosis.27
Maeda et al (2016) reported that, in TNBC, the coexpression of K5/6, AR, and p53 identifies high-risk patients. K5/6 upregulation may stabilize the cytoskeleton during invasion, whereas AR loss is associated with high mutant TP53 mRNA expression. Mutant p53 accumulation promotes chemoresistance and tumor survival. Together, these results indicate that tumors with low K5/6, low AR, and high p53 levels have the worst prognosis, whereas low K5/6, high AR, and high p53 levels predict better outcomes.28
Voutilainen et al (2021) reported that, in metaplastic BC (MpBC), high K5/6 expression is indicative of a basal-like phenotype associated with mutations in the PI3K/AKT pathway. This drives aggressive behavior and chemoresistance, resulting in a poor prognosis for basal-like MpBCs (K5/6/EGFR+). In contrast, nonbasal-like MpBCs (K5/6/EGFR–) did not relapse.29
Martin-Castillo et al (2015) demonstrated that high K5/6 expression contributes to trastuzumab resistance via persistent PI3K/AKT signaling in HER2+ basal-HER2 hybrid tumors. This pathway underpins poor prognosis and therapy resistance in K5/6+ HER2+ patients.30
Lee et al (2024) reported that increased K15 expression activates pathways involving FOXM1 (cell cycle), PKC/AP-1 (differentiation), and stem cell maintenance in Luminal B HER2-BC. K15 acts as a progenitor marker linked to self-renewal and basal-like signatures, enhancing migration, invasion, and metastasis. Consequently, K15 positivity is associated with shorter disease-free survival (DFS) and BC-specific survival (BCSS), indicating a poor prognosis.31
Merkin et al (2017) reported that, in TNBC and ER-/HER2– tumors, high K17 expression promotes nuclear p27 degradation, leading to cell cycle progression. By acting as a shuttle for p27, K17 promotes proliferation and coordinates with K16/K14 to disrupt the cytoskeleton synergistically, thereby enhancing invasion and metastasis. High K17 expression is associated with a poor prognosis and worse 5-year event-free survival (EFS) in advanced receptor-negative tumors.32
Hendawy et al (2023) reported that increased K18 expression after NAC reflects apoptosis in luminal HER2– BC, as cleaved K18 fragments circulate in the blood. While high K18 posttreatment indicates a better response to treatment, elevated K18 pretreatment is associated with a poorer prognosis owing to its influence on cell adhesion, motility, and aggressive tumor biology.33
Morales et al (2018) reported that, in luminal, TNBC and HER2+ BC, upregulation of KRT19 drives EMT by reducing E-cadherin and activating the PI3K/AKT/mTOR survival pathways. This promotes chemoresistance through upregulation of the stress response. Fragmented K19 disrupts the cytoskeleton, increasing invasion and metastasis. High CTC-KRT19+ levels are associated with shorter progression-free survival (PFS) and overall survival (OS), indicating a poor prognosis.34
Wang et al (2019) demonstrated that, in luminal and HER2+ cancers, the expression of KRT19 in CTCs is associated with metastatic potential and EMT activation. High CTC-KRT19 levels are associated with shorter distant metastasis-free survival (DMFS) and OS, whereas a reduction after treatment is associated with better PFS.35
Tang et al (2022) demonstrated that, in HER2-high BC, low KRT17 expression is associated with a poor prognosis. KRT17 plays a role in keratinization, cytokine interactions, and IL-17 signaling. Its loss promotes immune evasion, increasing migration, and metastasis.36
Zhao et al (2019) reported that, in luminal BC, which is prone to bone metastasis, high KRT23 expression is part of a bone metastasis gene signature. This indicates a greater risk of metastasis and worse outcomes.37
In Tables 2–4, bridging preclinical findings to clinical applications demands further research shown in Figures 2 and 3.

EGFR/MAPK/PI3K/AKT (12 studies) and Wnt/β-catenin (10 studies) are the most frequently investigated pathways, primarily in human and in vitro studies. Pathways such as EMT and FOXM1/PKC/AP-1 are also prominent but lack in vivo validation.
K5/6 dominated research (25 studies), especially in TNBC/basal-like subtypes, followed by K19 (12 studies), K14 (10 studies), K8/18 (8 studies), and K17 (6 studies). Most evidence comes from human studies, with 18, 9, 7, 5, and 4 studies identified for K5/6, K19, K14, K8/18, and K17, respectively.

Discussion

Keratin Signaling Pathways and Tumor Progression Mechanism
Keratin, the largest and most diverse class among IFs, serves as a critical marker for proteins in BC.38 Since most clinical evidence supporting keratin as a biomarker relies on IHC-based methods detection,27,28,30 keratin expression in BC has mainly been assessed using antibody-based assays as lineage-defining,5,26 prognostic, and therapy-stratifying markers.28 Across the included human studies, keratins such as K5/6, K14, K17, and K8/18 were mainly evaluated using semi-quantitative IHC scoring systems based on staining intensity and/or the proportion of positively stained tumor cells. These were often divided into high- and low-expression groups. However, differences in antibody clones, antigen retrieval methods, scoring thresholds, and inter-observer interpretation across studies may affect reproducibility and limit direct comparability, which is an important consideration for clinical application. In BC, keratin proteins play pivotal roles in tumor progression, with specific isoforms driving distinct signaling pathways that influence tumor behavior and therapeutic responses. For example, in K5/6+ tumors, the main pathway involved is the EGFR/MAPK/PI3K/AKT axis, especially in basal-like TNBC, where it promotes growth, survival, and metastasis.27 These cancers often have mutant p53, which makes them resistant to drugs.28,39 These findings emphasize the potential of keratin-specific pathway targeting for personalized therapies.
Keratin subtypes play pivotal roles in regulating critical signaling pathways that are involved in the progression of BC. These subtypes have been found to modulate tumor mechanisms, including metastasis, chemoresistance, immune evasion, and stemness, among others. This review systematically mapped the entire pathway into five major pathways across 79 studies shown in Figure 2. These pathways have been reported to be involved in keratin-associated signaling mechanisms in a variety of experimental studies, including in vitro, in vivo, and clinical studies in humans.
The consistent correlation between keratin expression and core oncogenic pathways suggests that individual keratin subtypes could serve as biomarkers associated with particular pathways,27,28,30 helping to identify tumors that depend on different signaling axes. This is consistent with translational techniques in which biomarkers can expedite drug development by classifying patients based on key pathway vulnerabilities rather than histology alone.40

Role of Keratin 5/6 as Prognostic Marker
K5/6 are basal keratins expressed in myoepithelial and progenitor cells and are associated with aggressive tumor phenotypes, including high-grade tumors, lymphovascular invasion, and chemoresistance,26,39 via several signaling pathways,26,27,39,41–43 as shown in Figure 4.

K5/6 EGFR is a receptor tyrosine kinase that is activated by various ligands, including EGF or TGF-α. In patients with TNBC and BLBC, excess EGFR expression is frequently detected.44 Moreover, the MAPK pathway is characterized by a cascade of events that begin with the activation of EGFR, which is followed by the activation of RAS. Subsequently, RAF is activated, followed by the activation of MEK. The final activation is that of ERK, which translocates into the nucleus to activate the transcription of genes involved in proliferation and metastasis.45 Moreover, the PI3K/AKT pathway is also activated downstream of EGFR. Upon activation, EGFR stimulates PI3K to generate PIP3, which serves as a docking site for the activation of the AKT serine/threonine kinase.46 The activation of AKT has been demonstrated to promote cell survival by inhibiting apoptosis. This pathway plays a pivotal role in the progression of tumors, specifically in metastasis and chemoresistance. On the other hand, activation of the EGFR/MAPK/PI3K axis alters the dynamics of the cytoskeleton, thereby increasing cell motility. MAPK/ERK signaling activates pro-EMT transcription factors, including Twist, Snail, and Zeb1, which in turn lead to the suppression of E-cadherin and the upregulation of N-cadherin and vimentin. Concurrently, PI3K/AKT has been shown to increase the expression of MMP-2 and MMP-9, thereby promoting ECM degradation and facilitating invasion. Furthermore, AKT activation has been demonstrated to increase VEGF expression, thereby promoting angiogenesis and contributing to destabilization of the vascular endothelium.47 The mutant p53 protein has been shown to increase the activation of EGFR and the PI3K/AKT pathway, thereby increasing the resistance of cancer cells to chemotherapy agents such as anthracyclines or platinum-based drugs. Furthermore, mutant p53 disrupts HRR while activating alternative DNA repair pathways, thereby reducing tumor cell dependency on PARP and contributing to PARP inhibitor resistance.48
From a translational standpoint, the intense correlation of K5/6 expression with chronic EGFR/MAPK/PI3K/AKT activation in basal-like TNBC elevates K5/6 beyond that of a prognostic marker to that of a tumor subgroup that necessitates active signaling pathways reliance.27,28,30 Consequently, K5/6 expression may serve as an adjunct biomarker associated with tumors that are largely dependent on EGFR- or PI3K-driven survival signaling and therefore supportive of pathway-targeted or combination therapies.40

Role of Keratin 19 as Prognostic Marker
K19 in cancer progression, highlighting its dual function as an epithelial marker for detecting CTCs and micrometastases and its involvement in key oncogenic pathways, as shown in Figure 5.

This figure outlines two critical pathways.17,33,34,49–51 First, Wnt ligands (eg, Wnt3a) bind to the Frizzled receptor, thereby inhibiting the β-catenin destruction complex (which consists of GSK3β, APC, and Axin). Consequently, β-catenin accumulates in the cytoplasm and subsequently translocate into the nucleus, where it binds to TCF/LEF transcription factors to activate protumorigenic target genes, including c-Myc, Cyclin D1, and various MMPs.52 Second, the activation of the Notch receptor is initiated by the binding of ligands, such as Jagged or Delta-like proteins, resulting in the release of the NICD. The NICD translocates into the nucleus and activates genes, including Hes1 and Hey1, which play key roles in maintaining stemness and promoting EMT.53 During the metastatic process, a decrease in E-cadherin and an increase in vimentin serve as hallmarks of EMT. Nuclear β-catenin has been shown to induce the expression of transcription factors such as Snail, Twist, and Zeb1, which have been demonstrated to suppress CDH1 (the E-cadherin gene) and upregulate vimentin expression. Moreover, Notch signaling has been shown to increase the expression of MMP-2 and MMP-9, which in turn degrade the ECM and facilitate tumor cell invasion.54 K19+ cells that also exhibit active Wnt/Notch signaling have been shown to exhibit disrupted polarity and heightened motility, which renders them more susceptible to entering blood vessels or lymphatics.55 Moreover, the activation of the PI3K/AKT pathway by K19 fragments (eg, CYFRA 21–1) further supports chemoresistance by upregulating antiapoptotic proteins such as Bcl-2 and Survivin. Furthermore, nuclear β-catenin has been shown to increase the expression of ABC transporters, such as P-glycoprotein, which play a pivotal role in the active efflux of chemotherapeutic agents, thereby contributing to multidrug resistance. Furthermore, K19+ cells exhibiting active Notch signaling have been shown to acquire characteristics of CSCs, thereby conferring enhanced resistance to chemotherapy and radiation therapy.56
Likewise, K19-driven crosstalk across Wnt/β-catenin, Notch, and PI3K/AKT signaling determines a keratin-associated tumor state with defined features of stemness, epithelial–mesenchymal transition, and therapeutic resistance.17,18,34,35 These features are consistent with the biomarker–pathway axes commonly employed in companion diagnostic approaches, which suggest K19-positive tumors are a biologically defined subgroup that can be targeted towards pathway-based therapeutic modulation rather than uniform cytotoxic therapy.40

Role of Keratin 8/18 as Prognostic Marker
K8/18 is a luminal keratin marker with context-dependent functions in BC progression. In luminal BC, K8+ lymph node metastases correlate with better survival, whereas loss of K8/18 in primary lymph node tumors is associated with EMT and worse outcomes. In TNBC, cytoplasmic K8/18 promotes proliferation and poor prognosis by activating the PI3K/AKT/ERK pathways and its plasminogen receptor function enhances invasiveness, as shown in Figure 6.

K8/18 plays a multifaceted role in BC progression.16,24,49,57,58 Initially, PI3K is activated by receptor tyrosine kinases or the ER. Once activated, it converts PIP2 into PIP3, which in turn serves as a docking site for AKT recruitment and activation.59 Second, AKT is phosphorylated by PDK1 and mTORC2 following binding to PIP3. The substance plays a pivotal role in the suppression of apoptosis and the promotion of cell proliferation through several mechanisms, including the inhibition of proapoptotic proteins (eg, BAD and caspase-9) and the activation of transcription factors (eg, NF-κB).60 Third, ERK is part of the MAPK pathway, but in this context, it is activated through crosstalk with PI3K/AKT signaling. ERK has been demonstrated to regulate the expression of genes associated with cell cycle progression (eg, Cyclin D1) and cell motility.61 Via this pathway, the process of invasion is initiated via apoptosis inhibition and cell cycle progression. AKT has been shown to phosphorylate and inactivate proapoptotic proteins, such as BAD and FoxO, thus enabling cancer cells to survive despite DNA damage or microenvironmental stress. K8/18 has the capacity to reinforce this effect by stabilizing the PI3K/AKT complex. Furthermore, AKT activates mTOR, thereby enhancing protein synthesis and cell growth, whereas ERK increases Cyclin D1 expression, accelerating the G1-to-S phase transition. The result of the process under investigation is uncontrolled cell proliferation.62,63 Second, AKT and ERK have been demonstrated to increase the expression of MMPs, including MMP-9, which are involved in the degradation of the ECM, thereby facilitating tumor invasion.64,65 The PI3K/AKT pathway has also been demonstrated to activate NF-κB, thereby increasing the expression of adhesion molecules (eg, ICAM-1) that promote tumor cell interactions with the vascular endothelium.66,67 ERK further promotes partial EMT by downregulating E-cadherin, contributing to a more invasive phenotype.68,69

Role of Keratin 14 as Prognostic Marker
K14 contributes to BC progression, metastasis, and aggressive tumor behavior. K14 expression, driven by H3K27me3/SPI repression, promotes invasion through the Rho/ROCK/integrin pathway, as shown in Figure 7.

K14 significantly influences pivotal oncogenic pathways,9,11,12,43 including (a) Rho GTPases (eg, RhoA, Rac1, and Cdc42) are signaling proteins that play essential roles in the organization of the actin cytoskeleton and in cell motility. The activation of these processes is initiated by surface receptors, which include integrins and G protein–coupled receptors (GPCRs).70 (2) ROCK is a significant effector downstream of RhoA that phosphorylates substrates such as myosin II regulatory light chain (MLC), thereby promoting cytoskeletal contractility.71 (3) Integrins are cell adhesion receptors that link the ECM to the cytoskeleton.72 In this pathway, K14-driven tumor progression is initiated via invasion and immune evasion. Invasion via cytokeratin and cytoskeletal reorganization.73 The use of RhoA/ROCK signaling triggers MLC phosphorylation, thereby increasing actomyosin contractility and promoting the formation of stress fibers and invadopodia. These are cellular protrusions that degrade the ECM during cancer cell invasion.74,75 Integrins have been shown to activate FAK/Src pathways, which in turn promote invadopodia formation and the release of MMPs, thereby facilitating the breakdown of the ECM.76,77 Tumor immune evasion is generally associated with infiltration of Tregs and increased production of immunosuppressive cytokines (eg, TGF-β, IL-10).78 Furthermore, activation of Rho/ROCK signaling has been implicated in cytoskeletal remodeling and invasive behavior of tumor cells.71 Integrin αvβ3 signaling has been associated with formation of immunosuppressive tumor microenvironments,79 such as recruitment and functional polarization of regulatory immune cells.80 This phenomenon is facilitated by activated STAT3, which functions as an inducer of PD-L1 expression and integrin-mediated tumor cell adhesion to immune cells. The resulting effects include the inhibition of cytotoxic T-cell function and the promotion of immune evasion.81,82
In addition to its prognostic value, K14-mediated activation of Rho/ROCK and integrin signaling highlights a potential vulnerability of K14-high tumors to interventions targeting the cytoskeleton and mechanotransduction, particularly in invasive and immune-evasive disease states.11,12,25 This concept aligns with the increasing focus on reprogramming tumor ecosystems instead of causing direct cell death by targeting cytoskeletal dynamics, stromal interaction, and immune response modulation.83

Role of Keratin 17 as Prognostic Marker
K17 expression is correlated with poor prognosis in aggressive BC subtypes, particularly TNBC and basal-like tumors, as shown in Figure 8.

K17 drives tumor progression through multiple pathways.9,32,36 The IL-17/cytokine receptor pathway is a proinflammatory signaling pathway that involves the cytokine IL-17 and its receptor, IL-17R. IL-17 is produced by Th17 cells, mast cells, or even tumor cells themselves. The substance in question has been demonstrated to bind to IL-17R, which is expressed on the surface of both tumor cells and stromal cells.84,85 Upon binding, IL‑17R activates NF‑κB and MAPK, which in turn upregulates the expression of proinflammatory genes such as IL‑6 and MMP‑9.86 Furthermore, the activation of STAT3 results in increased tumor cell survival.87 IL-17 has been demonstrated to increase the production of MMP-9, which in turn degrades the ECM, thus facilitating tumor invasion. IL-17 has been shown to induce the release of TNF-α and IL-6, thereby creating a prometastatic microenvironment.88 IL-17 signaling induces the recruitment of myeloid-derived suppressor cells (MDSCs), which have been shown to suppress antitumor immunity and promote metastasis.89 Furthermore, the phenomenon of chemoresistance, which is induced by crosstalk with EGFR, can be attributed to the activation of antiapoptotic pathways via the same mechanism.90 In addition, IL-17 indirectly activates EGFR through the release of its ligands, including amphiregulin.91 EGFR activation has been demonstrated to stimulate the PI3K/AKT and MAPK pathways, thereby increasing cell survival.92 The resistance mechanism is initiated when STAT3, following activation by IL-17, results in the upregulation of the expression of Bcl-2 and multidrug resistance 1 (MDR1), thereby increasing the resistance of tumor cells to chemotherapy.93,94
On a more specific level, K17-driven activation of the IL-17-STAT3 inflammatory pathway mediates the association between keratin expression and cytokine-mediated survival, immune suppression, and chemoresistance in aggressive subtypes of BC.37,40 This context for inflammatory signaling is consistent with treatment paradigms driven by an ecosystem orientation that seek to break pro-tumor immune-cytokine loops.83
Significantly, many poor-prognosis keratin subtypes coalesce around inflammation-survival signaling networks primarily focused on NF-κB activation. NF-κB encapsulates epithelial-mesenchymal transition, immunological modulation, and apoptotic resistance through interactions with PI3K/AKT signaling, cytokine pathways, and cytoskeletal remodeling, enhancing aggressive tumor phenotypes.15,34 This is in line with data from other cancers that demonstrate that NF-κB-based programs favor invasion, survival, and therapy resistance, suggesting a valuable connection between keratin-driven phenotypes and stress-adaptive oncogenic modulation.95
While this review focuses primarily on individual keratin subtypes for convenience of interpretation, several studies from the dataset also indicate the biological role and clinical impact of multi-keratin expression. Examples encompass the combination of K5/6 with androgen receptor and p53 status in high risk TNBC,28 the combined expression profiling of K5, K6, K14, and K17,11 in stemness-associated tumor programs, as well as the K14-luminal (K18) balance regulated by ECM.23 These results suggest that function of keratin in BC is a result of the interaction of keratin networks and not a result of single markers by itself, thus are dynamic characteristics of tumor cell identity. While spatial transcriptomics and single-cell profiling were not used directly in the studies referenced, emerging data from these strategies are expected to optimize multi-keratin signatures via enhanced resolution of intratumoral heterogeneity, potentially gaining future utility for risk stratification and precision oncology.

Conclusion
This review confirms that keratin in BC not only acts as a diagnostic and prognostic biomarker but also plays an active biological role in tumor progression. Its natural functions include maintaining cytoskeletal stability and cell adhesion. However, specific subtypes, such as K5/6, K14, K17, K19, and K8/18, were selected because they are the most common and clinically significant in BC, have the most published evidence, and shown to promote oncogenic pathways. These include EGFR/MAPK/PI3K/AKT, Wnt/β-catenin, Rho/ROCK/integrin, and Notch. These pathways drive invasion, EMT, metastasis, chemoresistance, and immune evasion. Consequently, any therapeutic approach targeting keratin must meticulously consider the risk of disrupting its essential structural roles in normal tissues.
Indeed, the subtype-specific importance of keratin expression is especially pronounced in triple-negative and basal-like BC, with the most impactful prognostic value due to keratin classification. Basal keratins K5/6, K14, and K17 consistently identify TNBC and basal-like tumors and are associated with aggressive phenotypes characterized by EMT, invasion, metastasis, immune evasion, and chemotherapy resistance. On the other hand, K19 is less subtype specific and mainly serves as a marker of tumor dissemination, with circulating tumor cells in luminal, HER2-positive, and TNBC subtypes exhibiting metastatic tendency and poor clinical outcomes. Keratin 8/18 has a context-sensitive function as a marker of epithelial integrity and treatment response in luminal BC and promotes proliferation and aggressiveness through PI3K/AKT/ERK signaling when expressed aberrantly in TNBC. Overall, this approach suggests that keratin profiling is most useful when interpreted within specific BC subtypes, especially TNBC, rather than being applied uniformly across all molecular categories.
It is recommended that future research efforts concentrate on identifying keratin subtypes that exhibit specific dominant expression in tumor cells while displaying minimal or redundant functions in healthy epithelium. Examples of such keratins include K5/6 in basal-like TNBC and K14/K17 in tumors with proven invasive characteristics. These subtypes are more rational candidates for functional inhibition or pathway-specific knockdown. For subtypes whose natural functions are critical to tissue integrity, exploiting their overexpression as a targeting marker for precision drug delivery systems, such as PDC-K1-targeted therapy, may be safer. This dual strategy is designed to achieve a balance between the potential to suppress oncogenic signaling and the minimization of harm to normal tissue homeostasis.