Introduction

Inflammation is now recognized as playing an important role in different stages of tumorigenesis, including initiation, promotion, malignant conversion, invasion, and metastasis, and NF-κB is one of the major factors linking inflammation and cancer [1]. Multiple observations have highlighted the aberrant or constitutive NF-κB activation in a number of human cancers, including lymphoma, liver, lung and breast cancers. Abnormal NF-κB activation is also driven by environmental stimuli commonly associated with carcinogenesis, such as tobacco and/or alcohol use, and irradiation. The major tumorigenic function of NF-κB has been linked to disturbed regulation of the transcription of targets associated with the cell cycle including cyclin D1/D2 and CDK 2/CDK6, and apoptosis, including cIAP1, XIAP and c-FLIP, resulting in abnormal cancer cell progression and the suppression of apoptosis respectively [2]. NF-κB activation was also reported to be involved in tumorigenic angiogenesis and tumor cell invasion [3]. Constitutively active NF-κB signaling results in secretion of major inflammatory cytokines or chemokines, including TNFα, IL-1 or IL-6, which, through a positive feedback loop, increase NF-κB activation, further contributing to uncontrolled growth and malignant transformation. Therefore, a better understanding of NF-κB and its association with tumor-promoting inflammation and anti-tumor immune suppression will likely facilitate the development and optimization of cancer prevention and treatment.

Epithelial to mesenchymal transition (EMT) is a process in which epithelial cells acquire mesenchymal phenotype and lose epithelial features. EMT involves a sequence of steps that include 1) loss of stable epithelial cell–cell junctions, 2) loss of apical–basal polarity and interactions with basement membrane, 3) cytoskeletal rearrangements leading to acquirement of fibroblast-like morphology and cytoarchitecture, 4) increased migratory capacity and 5) acquirement of invasive properties. EMT normally occurs during early embryonic development or during wound healing process in adults. EMT is also activated during carcinogenesis and is involved in cancerous expansion, metastasis, cancer recurrence and development of several types of fibrosis. It is important to emphasize that EMT is associated with phenotypic heterogeneity due to the often incomplete transition from epithelial to mesenchymal state, resulting in an array of intermediate states in which cells retain both epithelial and mesenchymal characteristics. These intermediate states are collectively named a state of epithelial-mesenchymal plasticity. The completion of EMT is typically accompanied by a switch in intermediate filament utilization from cytokeratins to vimentin. In the early 1990s, a number of transcription factors (TFs), including Slug, Snail, E47, Twist1, Zeb1 and Zeb2, were identified by means of their ability to induce EMT phenotypes and orchestrate the process. These EMT TFs control cell–cell adhesion, cell migration and degradation of the extracellular matrix. It also became apparent that activation and execution of EMT does not require permanent changes in DNA sequence and instead is fine-tuned by epigenetic regulators. Given the heterogeneity of EMT states and pleiotropy of observed intermediate phenotypes it has become clear that EMT state should be defined based on collective features including activity of core EMT TFs as well as morphological and cytological phenotypes [4]. In this review, in an effort to better understand how NF-κB-driven inflammation contributes to carcinogenesis, we attempt to comprehensively summarize the current knowledge of the involvement of the NF-κB signaling in the control of core EMT changes including cytoskeleton remodeling, loss of apical–basal cell polarity, cell–cell adhesion weakening and cell–matrix adhesion remodeling, acquisition of cell motility and basement membrane invasion. In this review, we approach the subject from cancer cell-centric view to highlight the role of NF-κB signaling in cancer cells undergoing transition. While NF-κB signaling plays an important role in modulating tumor microenvironment (reviewed in [5]), this is beyond the scope of this summary.

NF-κB structure and pathway overview

The NF-κB family and structure

The Nuclear Factor-κB (NF-κB) family of transcription factors regulates a large number of genes involved in a multitude of functions, including cell survival, proliferation, and immune responses. This family consists of five proteins—p65 (RelA), RelB, c-Rel, p50/p105 (NF-κB1), and p52/p100 (NF-κB2), all of which contain a highly conserved 300-residue long region, termed Rel Homology Domain (RHD) responsible for dimerization, DNA recognition, DNA binding and nuclear localization [6, 7] (Fig. 1A). The NF-κB family members can be further divided into two subgroups based on the sequences C-terminal to the RHD. One group, consisting of RelA, RelB, and c-Rel, contains a transactivating C-terminal region, while the other group, consisting of p50/p105 and p52/p100, on its C-terminus contains a structural motif, the death domain (DD). The members of each group can form inter- and intra-, homo- and hetero-dimers. Depending on the presence or absence of the transactivating domain, they function as either activators or repressors of transcription [8, 9]. The formation and stability of the NF-κB dimers is dependent on the sequence of amino acid residues in direct contact with each other, forming the interface, while amino acid residues outside of the interface modulate the local binding environment. The most abundant form in most cells, p50:p65(RelA) heterodimer, is one of the most stable dimers, whereas RelB homodimer does not exist in vivo due to the low stability of the RelB dimerization domain destabilized by non-interfacial amino acid residue interactions [10,11,12]. The first x-ray crystal structure of NF-κB p50 homodimers bound to DNA, resolved by Harrison and Sigler, showed that the RHD folds into two immunoglobulin-like domains [13, 14]. The N-terminus of one of the domains spans 160–200 amino acids and interacts with the major grove of DNA in base-specific manner, while the C-terminus of the other domain, being about one hundred amino acids in length, contributes to the hydrophobic residue-mediated dimerization, while interacting with DNA in nonspecific manner. Resolved NF-κB p50 homodimer-DNA complex provides evidence that the entire RHD scaffolding is required for the DNA recognition and interaction [6, 12, 15, 16]. The NF-κB dimers that translocate to the nucleus bind to decametric DNA sequence motifs containing the general consensus—GGGRNNYYCC (N denotes any nucleotide, R is for purine bases, and Y is for pyrimidine bases), known as κB sites. The p50 and p52 proteins prefer κB sites comprised of two GGGRN half-sites, separated by A/T base pair. The RelA, c-Rel and RelB proteins bind to κB sites containing two YYCC half-sites. The heterodimers (p50:RelA or p50/p52:RelB) show similar binding affinities to both types of κB sites. These mechanisms allow each hetero- or homodimer to mediate discrete cellular responses dependent on physiological contexts in response to numerous stimuli [10, 15, 17]. Importantly, however, NF-κBs are also able to bind to κB DNA sites with significant deviations from the consensus sequences. If the deviation occurs within the central region of the consensus sequences, the overall binding conformation remains the same, given the flexibility of the linker region, but the stability or the binding affinity, may change [12]. These unique features of NF-κBs structure and DNA binding ability allow NF-κB regulating the numerous genes and processes.

Fig. 1
figure 1

Schematic representation of NF-κB, Iκß, and IKK family members. (A) NF-κB family members share the RHD (Rel Homology Domain), important for DNA binding and dimerization. The functional domains of each subunit are indicated schematicaly: TAD = transcription activation domain; LZ = leucine zipper; GRR = glycine-rich domain; ANK = ankyrin repeats DD = death domain (B) Iκß family members share ANK domain that allows interaction with the RHD of NF-κB. Other indicated domains include PEST = proline/glutamic acid/serine/threonine-rich sequence. (C) The three IKK subunits are represented with domains that typify each protien: HLH = helix-loop-helix; NBD = NEMO-binding domian; CC = coiled-coil; ZF = zinc finger

Full size image

The IκB family and structure

In the absence of stimuli, NF-κB is normally sequestered in the cytosol through the interactions with the proteins of the IκB family. IκB proteins are a subfamily within the large Ankyrin Repeat Domain (ARD) containing superfamily and can be classified into three categories: classical IκB (IκBα, IκBβ, and IκBε), NF-κB precursors (NF-κB precursor p105 and p100), and nuclear IκB (IκBζ, Bcl-3 and IκBNS). Classical IκBs contain phosphorylation and poly-ubiquitination sites at the N-terminus and the NF-κB:DNA complex disrupting PEST region, composed of proline, glutamic acid, serine, and threonine, at the C-terminus (Fig. 1B). IκBα contains a nuclear export signal that mediates the localization of the NF-κB:IκBα complex to the cytoplasm [12]. The X-ray crystal structure of IκBα:NF-κB p50/65 heterodimer complex shows the conserved mode of IκB binding: the ankyrin repeat of the IκBα runs in the antiparallel direction, curves towards, and binds to the NF-κB heterodimer in a “cupped hand” manner, inhibiting its binding to DNA. The ankyrin repeats 1 and 2 form hydrophobic contact with the nuclear localization signal located at the dimer’s C-terminal domain, while ankyrin repeats interact with the RHD at the same terminus. The acidic property of the ankyrin repeats and PEST region repels the positively charged N-terminal domain of the p65 subunit, which undergoes significant conformation change into a “locked” form that completely masks the nuclear localization signal. A similar binding pattern is observed in IκBβ:NF-κB p50/65 heterodimer, but with less dependence on the interaction between the N-terminal domain of the NF-κB dimer. IκBβ does not directly bind to the p65 subunit N-terminal domain, leaving the nuclear localization signal and the DNA-binding domain free to bind to DNA. Consequently, studies have shown that the IκBβ:NF-κB p50/65 complex is found in both cytoplasm and nucleus, while the IκBα:NF-κB p50/65 complex is exclusively located in the cytoplasm [8, 16, 18]. A series of biophysical experiments, including single-molecule fluorescence resonance energy transfer (FRET), have shown that classical IκBs are inherently unstable and remain incompletely folded in their free states, subjected to steady-state, signal-independent degradation by 20 s proteasome and the C-terminal PEST of IκBs. Upon binding to NF-κB, IκBs switch from the extended to compact form and become stable until the PEST region is degraded in signal-dependent manner through phosphorylation of the N-terminal response domain [10, 19]. The nonclassical IκB proteins, NF-κB precursors, are similar to the classical IκB proteins in two major aspects: they contain ankyrin repeats and mediate the gene expression level through the interaction with NF-κB dimers. Unlike the classical IκB proteins that form 1:1 complex with the NF-κBs, the nonclassical IκB proteins are capable of binding to more than one NF-κB dimers through their oligomerization domain, forming a multimeric complex. They also show different binding affinities for the NF-κB members: classical IκBs bind to NF-κB dimers that contain at least one p65 or c-Rel subunits, while nonclassical IκBs are limited to binding p50 or 52 homodimers. Given the variation in the binding affinities, IκB proteins can function as modulators of NF-κB dimerization, determining the prevalence of the NF-κB dimers, which may play an important role in the NF-κB transcriptional specificity [12, 16]. Finally, nuclear IκB proteins do not contain the N-terminal signal-dependent phosphorylation sites, or the C-terminal PEST region, but they are still classified as IκB family, considering that they have ankyrin repeats and are capable of binding to NF-κB subunits, namely p50 homodimers only. They are known to play an important role in controlling gene expression and immune homeostasis, as for example some experiments have demonstrated that mice were not capable of producing IL-6 response to the LPS treatment in the absence of nuclear IκB proteins [12]. In sum, all three classes of IκBs add to the complexity of NF-κB transcriptional specificity.

The IKK complex

IKK family, comprised of IKKα (IKK1), IKKβ (IKK2), and IKKγ (also known as NEMO), functions as a converging point for the majority of the NF-κB activating signaling pathways. NEMO, the key regulatory non-enzymatic scaffold protein, is required for the catalytical subunits to fully gain the inducible kinase activity, although the exact oligomerization state of the IKK complex is poorly understood. The IKK catalytic subunits are organized into five distinct parts: the N-terminal kinase domain, followed by the ubiquitin-homology region, leucine zipper and helix-to-helix motifs in the center responsible for dimerization, and finally the serine-rich region and NEMO-binding site at the C-terminus (Fig. 1C). The regions outside of the kinase domain mediate recognition and exact positioning of the IκB substrates, as studies have shown that IKK complex loses its binding specificity in the absence of leucine zipper and helix-to-helix motif regions [12, 16, 20]. The catalytic subunits exist in dimeric structure, forming homo- or hetero-dimers that resemble a pair of scissors. The kinase domain of the two monomers is located far from each other, incapable of stimulating intradimer trans-autophosphorylation, while they interact and activate the kinase domain of the neighboring dimers through NEMO-mediated ubiquitin chain network. The mutagenesis studies of IKKβ suggest that the dimerization of the kinase domain is necessary for NEMO binding and recruitment of the IκBα substrate [18, 20, 21]. While multiple X-ray structures of the fragments of NEMO have been reported, the full-length protein structure of NEMO subunit remains unknown [18, 20]. Assembling the structures of the isolated domains shows that NEMO is composed of the symmetrical helical-shaped (except the zinc-finger region) dimers, containing two helices (HLX1 and HLX2), two coiled-coil domains (CC1 and CC2) in configuration HLX1, CC1, HLX2, CC2, followed by leucine-zipper domain and C-terminal zinc-finger (ZF) region (Fig. 1C). The first two regions, HLX1 and CC1, form the NEMO binding site for IKKα/β, and the ubiquitin-binding motif is located on the C-terminus and encompasses leucine-zipper and ZF regions. Chemical cross-linking and equilibrium sedimentation analyses suggest that NEMO dimers can interact with IKKβ homodimers, forming helix-shaped hetero-tetramer. The two NEMO molecules interact with each other at the N- and C-terminus, while the two IKKβ molecules interact with each NEMO molecule individually, without interacting with each other. Stoichiometrically, this hetero tetramer can bind to two IKKα and IKKβ molecules, contributing to the IKK trans-autophosphorylation [18, 20]. The NEMO plays a crucial role in the NF-κB cascade for its ability to recognize and bind to the poly-ubiquitinated sites (both N-terminal methionine-linked di-ubiquitin and lysine 63-linked polyubiquitin) on the proteins involved in the NF-κB activation, functioning as an adaptor linking the catalytic subunits and other receptor signaling molecules [18, 22].

The NF-κB activation cascade

In response to immune and stress stimuli, NF-κB becomes activated via two major pathways, canonical and noncanonical (Fig. 2). The canonical activation pathway involves signal-induced proteolysis of the IκBs, particularly IκBα, regulated by IκB kinases (IKKs). Upon activation by pro-inflammatory signals such as cytokines or pathogen-associated molecular patterns (PAMPs), IKKs phosphorylate the two serine residues in the N-terminal signal receiving domain of IκBs, leading to the poly-ubiquitination of the adjacent lysine residues. This results in degradation of IκBs in the proteasome and freeing of NF-κBs. The freed p50-containing NF-κB dimers, the most common form being p50:RelA and p50:c-Rel heterodimers, translocate to the nucleus and bind to their target promoter sites. The canonical pathway is known to play a critical role in regulating immune responses, including lymphocyte activation and differentiation, innate immunity, and inflammation [18, 23]. A selective set of differentiating and developmental stimuli, largely belonging to the tumor necrosis factor receptor (TNFR) superfamily, are known to activate the non-canonical pathway. It is characterized by the processing of the NF-κB precursor protein p100 through the phosphorylation of its C-terminal serine residues by NF-κB inducing kinase (NIK) and/or IKKα. Increasing evidence suggests the involvement of the DD of p100 in the processing, as its removal leads to constitutive processing. The processed p52 subunit then dimerizes with RelB to enter the nucleus to regulate transcription of genes involved in lymphoid organ development, B cell maturation, osteoclast differentiation and broadly autoimmune and inflammatory responses [1, 23, 24]. The activation of NF-κB through the canonical pathway is rapid but transient and is terminated by the NF-κB-mediated re-synthesis of IκB proteins, which disrupts the NF-κB:DNA binding and results in export of the transcription factors back to the cytosol. In contrast, non-canonical activation is slow due to its dependence on the ubiquitination-regulated stabilization of NIK [17, 20].

Fig. 2
figure 2

Activation of canonical and non-canonical NF-κB signaling. The activation of the canonical pathway is induced by proinflammatory cytokines (e.g., IL-1ß, TNFα) binding to their respective receptors on the cell surface. This triggers the activation of TAK1 (transforming growth factor ß-activated kinase 1), which in turn activates the IKK complex, consisting of the regulatory NEMO and the catalytic subunits IKKα and IKKß. The activated IKK complex then phosphorylates the IκB protein, triggering Iκß ubiquitination and proteasomal degradation. The classical NF-κB dimers are released and translocate to the nucleus to regulate gene expression. Unlike the canonical pathway, the non-canonical pathway is activated by a distinct set of stimuli, activating the TNFR superfamily, which results in the stabilization and accumulation of NIK kinase. Increased NIK protein level phosphorylates IKKα, which in turn phosphorylates p100 protein, leading to partial degradation and conversion into the active p52 subunit. The p52 subunit forms a heterodimer with RelB, which translocate to the nucleus and activates the transcription of target genes

Full size image

The receptor-induced signaling cascade

Several receptor-induced IKK activation cascades have been identified, in which TNFR-and toll-like receptor/interleukin-1 receptor (TLR/IL-1R) superfamily-induced activations have been extensively studied. The TNFR-induced IKK activation pathway begins with extracellular ligands binding to TNFR, which recruits TNFR-associated factors (TRAFs) directly or through adaptor proteins. TRAFs contain N-terminal RING finger domain, followed by ZF, and C-terminal CC and TRAF-C region. Typically, the N-terminal region is responsible for dimerization and mediates lysine 63-linked polyubiquitination, while C-terminal is involved in trimerization and interactions with the receptor and adaptor proteins. Each termini provides a scaffold for TRAF aggregation and higher-order oligomerization and locally concentrates all of the associated signaling proteins, which facilitate the autoubiquitination, polyubiquitination, and downstream signaling. cIAP1/2, recruited by the CC region of TRAFs, drives the polyubiquitination of multiple proteins, such as receptor-interacting serine/threonine-protein kinase 1 (RIPK1), NIK, TRAF2, which leads to recruitment of downstream proteins, including NEMO and ubiquitin ligase. The binding of NEMO to the ubiquitin chain complex initiates the IKK activation, either through inducing the conformational changes or positioning IKK to have it exposed to phosphorylation by upstream kinases in the complex [2, 18, 20, 21]. The TLR and IL-1 superfamily shares a common Toll/1L-1R (TIR) intracellular domain, activating overlap** downstream cellular signals. The primary adaptor protein recruited by the TIR domain is MyD88, a member of the DD superfamily. The death domain of MyD88 oligomerizes with the IL-1R-associated kinase (IRAK) family members, IRAK4, IRAK1 and IRAK2, to form a complex termed myddosome. The IRAK4 initiates the auto-phosphorylation of itself and facilitates the phosphorylation of the other IRAK members in the complex. Next, the phosphorylated IRAK1 and IRAK2 recruit TRAF6, a ubiquitin E3 ligase that catalyzes lysine 63-mediated autoubiquitination and polyubiquitination in the signaling pathway, inducing the IKK activation followed by phosphorylation and ubiquitination of IκBs resulting in activation of NF-κB [16, 18, 23].

The NF-κB role in EMT

The NF-κB signaling has been implicated in multiple aspects of oncogenesis, including pro-inflammatory signaling, cell differentiation, migration, and tissue remodeling. Previous research has demonstrated constitutive activity of NF-κB, or mutations in genes encoding upstream regulators of NF-κB, in a significant number of human cancers, especially those of immune cell origin, such as leukemias and lymphomas. Recently, studies further suggested that NF-κB plays an essential role in induction and maintenance of invasive phenotypes in cancer, including EMT and metastasis, however the detailed mechanisms underlying NF-κB links to EMT remain unclear. Therefore, herein we summarize the current understanding of the involvement of NF-κB in EMT (Fig. 3), as delineating this relationship has a potential to facilitate the development and optimization of therapeutic strategies in cancer.

Fig. 3
figure 3

Schematic diagram of the EMT transition stages. Epithelial-mesenchymal transition (EMT) is a dynamic process in which epithelial cells undergo a transition into a mesenchymal state, leading to changes in their morphology, function, and behavior. The early-stage cells display epithelial features: apical-basal polarity is present, epithelial-associated proteins are expressed, and tight and adherens junctions hold the cells together. EMT involves a sequence of steps that starts with the loss of stable epithelial cell–cell junction, leading to loss of cell polarity and adhesion. The following remodeling of the cytoskeleton results in extensive rearrangement of actin filaments and microtubules, with cells gaining mesenchymal-like morphology and cytoarchitecture. The overexpression of regulators of EMT, such as transcription factors Snail, Slug, Twist, and Zeb1/2, leads to changes in gene expression, activating those associated with mesenchymal cell characteristics, including N-cadherin, MMPs, and vimentin. The mesenchymal cells exhibit increased migratory capacity and acquire invasive behavior, allowing them to disseminate into surrounding tissues. The major steps of EMT are highlighted with specific link to NF-κB signaling outlined

Full size image

NF-κB and dissolution of cell–cell junctions

Research investigating the relationship between NF-κB and EMT-associated dissolution of intercellular junctions focuses mainly on epithelial cadherin (E-cadherin), not only because it is a major epithelial marker and its decreased expression is considered a major hallmark of EMT, but also due to its function as a transmembrane protein and a major epithelial calcium-dependent cell adhesion molecule. Homophilic binding between E-cadherins of adjacent cells forms the basis of the epithelial cell–cell contacts—adherens junctions (AJs), which, together with other molecules, form an adhesion junctional complex. There are several reports highlighting the link between NF-κB and E-cadherin in AJs. Tripathi et al. [25] demonstrated that Rho GTPase-activating protein (RhoGAP) - Deleted in Liver Cancer 1 (DLC1), the downregulation of which is associated with prostate carcinoma (PCA), stabilizes AJs in PCA cell lines through binding to E-cadherin and as such has an inhibitory effect on NF-κB activation. Solanas et al. [26] also found that NF-κB as well as transcriptional activator β-catenin, both associate with E-cadherin at AJs. This interaction stabilizes AJs and has an inhibitory effect on the transcriptional activity of NF-κB and β-catenin, suppressing the expression of various mesenchymal markers central to EMT. Kuphal et al. [27] found that the constitutive activation of NF-κB led to decreased expression of E-cadherin within malignant melanoma cells, leading to the concomitant increase in free cytoplasmic β-catenin further leading to the p38 MAPK-mediated activation of NF-κB. Zipper-interacting protein kinase (ZIPK) or Death-Associated Protein Kinase 3 (DAPK3), is a part of the death-associated protein kinase family regulating apoptosis. Li et al. [28] found that the elevated levels of ZIPK in gastric carcinoma (GC) cells are linked to increased expression of Snail and Slug, decreased expression of E-cadherin and overexpression of mesenchymal markers and dissolution of intercellular junctions. Furthermore, it was demonstrated that the increased activation of Akt mediated by ZIPK does not lead to increased activation of PI3K/Akt/GSK3β but rather the PI3K/Akt/ΙΚΚ/IκBα/NF-κB signaling axis, presumably leading to the significantly increased expression of Snail and Slug and induction of the downstream EMT phenotype. In another study, Gao et al. [29] investigated the role of insulin-like growth factor-binding protein 2 (IGFBP2) in pancreatic ductal adenocarcinoma (PDAC). They found that IGFBP2 overexpression resulted in significantly increased expression of Snail, decreased expression of E-cadherin at intercellular junctions, increased expression of mesenchymal markers, nuclear translocation and overactivation of NF-κB and dissolution of intercellular junctions. Furthermore, Gao et al. demonstrated that IGFBP2-induced EMT is dependent on the increased activation of NF-κB, which they found to be linked to increased activation of the PI3K/Akt/ΙΚΚ/IκBα/NF-κB signaling axis. Cichon and Radisky [30] looked closer at NF-κB signaling to elucidate the molecular mechanism underlying matrix metalloproteinase 3 (MMP3)-induced EMT. MMP3 was shown to induce EMT in mammary epithelial cells via increased expression of Rac1b, an activated splice variant of Rac1 Rho GTPase, and subsequent stimulation of ROS production. Cichon and Radisky verified that MMP-3/Rac1b/ROS induces EMT in mammary epithelial cells. Presence of MMP3 resulted in significantly increased expression of Snail, significant activation of a tumorigenic transcriptional profile, including alterations of transcripts related to intercellular adhesion, mesenchymal morphology, and dissolution of intercellular junctions. They determined that MMP3/Rac1b/ROS-induced EMT requires ROS-dependent activation of NF-κB and that the activation of NF-κB results in upregulation of Snail via direct binding of NF-κB to its promoter. Cheng et al. [31] determined the status of EMT transcription factors Twist, Zeb1, and Zeb2 alongside Snail as they investigated the previously established association of tumor hypoxia, the expression of hypoxia-inducible factor-1 (HIF-1) and the constitutive activation of NF-κB with the development of pancreatic cancer (PC). They found that hypoxic conditions or overexpression of HIF-1α led to increased NF-κB activity, resulting in upregulation of Twist but not Snail, Zeb1, or Zeb2. Although the findings of Cheng et al. continue to corroborate the general relation demonstrated by the findings of Li et al., Gao et al., and Cichon and Radisky, in which increased activation of NF-κB leads to an EMT phenotype including the dissolution of intercellular junctions, the findings of Cheng et al. provide nuance to the specific EMT transcription factor regulation by NF-κB, which can be cell type dependent. Indeed, Chua et al. [32] found that mammary epithelial cells treated with TNFα or transduced to overexpress a constitutively active form of the p65 subunit of NF-κB, undergo EMT driven by an increased expression of Zeb1 and Zeb2 but not Snail or Slug, further highlighting potential cell type specific context of the effect of NF-κB on EMT TFs. Besides the PI3K/Akt, the TGF-β1/Smad signaling pathway has also been implicated in the EMT-associated increased activation of NF-κB via the canonical ΙΚΚ/IκBα/NF-κB axis. Transforming growth factor-β1 (TGF-β1) is a key mediator of EMT that has been shown to induce decreased expression of E-cadherin, increased expression of mesenchymal markers, and gain of mesenchymal morphology highlighted by dissolution of intercellular junctions. Lee et al. [33] found that TGF-β1 treatment of breast cancer (BC) cells results in significant activation of IκBα and NF-κB, significant increases in Snail and Slug expression and significant decreases in levels of E-cadherin and development of EMT phenotype. In summary, these findings indicate that the dissolution of intercellular junctions, mainly through downregulation of E-cadherin, is mechanistically linked to increased NF-κB activity during EMT, both through increasing the translocation of NF-κB to the nucleus and/or by increasing the overall expression and/or activity of NF-κB (Fig. 3). Further research is required to detail the mechanistic nuances of the effect of NF-κB signaling on the stability of AJ in the cell type specific manner.

NF-κB and cytoskeletal reorganization

The acquisition of mesenchymal-like phenotype during EMT leads to enhanced migratory and invasive abilities of cancer cells, which are mediated by cytoskeletal reorganization (Fig. 3). The crucial role of the cytoskeleton in the EMT was first proposed by Shankar et al. [34], who demonstrated that the inhibition of cancer-associated proteins resulted in the reduction of actin dynamics. Further research extensively examined dynamic reorganizations of the cytoskeleton required for EMT. Loss or inhibition of components of the actin network, specifically AHNAK (desmoyokin), septin-9, Eukaryotic Translation Initiation Factor 4E (eIF4E), or alarmin S100A11 led to reduction of formation of podosomes, invadopodia, filopodia and lamellipodia, resulting in reduced migration and invasion, and a reversal of EMT. Dinicola et al. [35] showed that treatment with inositol led to inhibition of PI3K and phosphorylation of Akt, which negatively impacted NF-κB and Snail leading to increased levels of E-cadherin, redistribution of β-catenin and reduction of membrane protrusions and cell motility. Avci et al. [36] found that co-treatment of glioblastoma cells with an NF-κB inhibitor that inhibits TNFα-induced IκBα phosphorylation—BAY 11–7082 and alkylating agent Temozolomide resulted in significant reduction in cell viability, suppressed NF-κB signaling, and enhanced apoptosis via actin skeleton modulation (Fig. 4). Aksenova et al. [37] investigated the transcriptional effect of actin-binding protein alpha-actinin 4 (ACTN4) on the RelA subunit of NF-κB. It was found that ACTN4 overexpression leads to co-activation of RelA, upregulation of matrix metalloproteinases MMP3 and MMP1, and enhancement of cellular motility. Zhao et al. [38] found that ACTN4 promotes expression of NF-κB target genes such as IL-1β and IL-8. In sum, NF-κB activity was shown to be central for cellular motility and invasiveness. Additionally, three major Rho GTPases – Rac1, RhoA and Cdc42, central for regulation of actin polymerization in cells, were found to be required for NF-κB transcriptional activity and pathway activation [39]. Cuadrado et al. [40] reported that Rac1 activates both the nuclear factor-like 2 (NRF2) pathway and NF-κB activity, indicating that Rac1 may also influence inflammation by coordinating activity of NF-κB and NRF2 transcription factors. The RhoA–NF-κB interaction has been shown to be important in cytokine-activated NF-κB processes, such as those induced by tumor necrosis factor α (TNFα), whereas Rac1 is important for activating the NF-κB response downstream of integrins. Detailed involvement of Rho-GTPases in NF-κB signaling is reviewed in Tong et al. [41]. Homeostasis of the cytoskeleton depends on balanced interactions between its filamentous components—actin filaments, intermediate filaments, and microtubules. Intermediate filaments support the plasma membrane and help maintain cell shape. During EMT, intermediate filaments become vimentin enriched [

Conclusions

Tumor promoting inflammation was recognized as a hallmark of cancer in 2011 [122], and the critical involvement of smoldering inflammation in carcinogenesis has been increasingly acknowledged since then Phenotypic plasticity was added to the list of hallmarks in 2022 [123] and involvement of major inflammatory pathway—NF-κB signaling—in regulation of phenotype plasticity has been now widely recognized. EMT is the developmental program that decreases cell–cell adherence allowing cells to acquire migratory properties and features of stem cell-like plasticity both contributing to acquired invasive properties, elevated metastatic and survival potentials. As summarized in this review it seems apparent that NF-κB-induced inflammation is a potent inducer, contributor and regulator of EMT phenotypes. Mechanistic links between NF-κB signaling and steps leading to transition from epithelial to mesenchymal phenotypes exist, but require further detailed mechanistic elucidation to reveal all the involved factors and uncover potential new therapeutic targets. In our search for therapeutics efficiently targeting EMT, it remains imperative to decipher both intracellular mechanisms induced by and driving inflammatory signaling and ensure interpretation of their role acknowledging proper microenvironmental context. Given the available evidence it is reasonable to infer that targeting NF-κB signaling may represent a valuable strategy to target EMT and related mechanisms in cancer.