Phase separations in oncogenesis, tumor progressions and metastasis: a glance from hallmarks of cancer

Phase separation is defined as the spontaneous aggregation of molecules when their concentration exceeds a certain threshold, thus forming a membrane-less compartment [21]. Typically, the interactions between macromolecules in LLPS are typically non-covalent and of low affinity [22, 23]. This process is often driven by the modification of intrinsically disordered regions (IDRs) within proteins [24, 25]. The concept of LLPS was first introduced in the biochemical field of biochemistry in 2009 by Hyman and colleagues with various milestone events followed subsequently (Fig. 1), offering a novel perspective on various MLOs distributed in cells (Fig. 2) [26]. Although several in silico tools help forecast the potential of phase-separated molecules (Table 1), comprehensive summaries of the characteristics and conditions that induce LLPS are limited.

The concept of a driver (or scaffold)/client is widely accepted. Proteins, DNA, and RNA can also be used as scaffolds. With multiple binding sites, these macromolecules facilitate weak interactions and trigger LLPS. The detailed structures are summarized below.

In this section, we focus on the conditions and the post-translational modifications (PTMs) which play a crucial role in regulating the dynamic transitions of molecules within the cell.

The interplay of various intracellular conditions, such as the concentration of proteins, pH level, and changes of the cellular milieu, alter the strength of polyvalent interactions. These conditions are key regulators of transitions within the cell. Furthermore, the concentrations of macromolecules are critical. When the concentration exceeds a critical threshold, the interaction between these macromolecules outweighs the forces that maintain homogeneity of the system, making the solution susceptible to phase separation. Conversely, when the concentrations are below this threshold, the components remain evenly distributed [89, 90]. The alterations of pH value can significantly impact LLPS by changing the surface charges of amino acids, the α-carbonyl groups, and the α-amino terminal protonation status. pH alterations affect the stability of specific proteins and change the secondary structure from ordered to disordered. Altering the protonation of amino acids directly influences the chemical properties of macromolecules, further altering their intermolecular interactions and triggering LLPS. For example, the decreased cytoplasmic pH, induced by external stimuli, can promote LLPS of naturally disordered proteins, as observed with Sup35 in yeast cells [91]. The increase in salt concentration and the addition of substances such as PEG3000 and glycerol can also modulate LLPS [73, 92]. Additionally, weak electrostatic interactions, driven by IDRs, are highly sensitive to changes in pH and ionic strength, potentially explaining LLPS induction due to environmental changes [17, 93]. In addition, temperature and stress levels can also trigger or disrupt LLPS by affecting the solubility of macromolecules [11]. Moreover, prion-like domains in proteins can sense pressure, influencing the solubility and phase behavior [94, 95].

The PTMs are crucial in the regulation of phase transitions by altering molecular interactions or directly modifying the potency of BMCs [96‐98]. PTMs can induce changes of biomolecules in the spatial structures and state of proteins [96, 99]. PTMs of RNA-binding proteins (RBPs) can directly weaken or enhance the interactions between components, contributing to the formation of RNP granules, serving as an example of an MLO that is composed of RBP and RNA [96]. PTMs can promote or inhibit polyvalent interactions by influencing the condition of proteins, thus affecting the occurrence of LLPS [100]. Notably, the Lys residues within the IDRs are particularly prone to get SUMOylation, a modification that significantly contributes to the formation of the promyelocytic leukemia nuclear bodies (NBs). De-SUMOylation can lead to the release of a constituent protein and the separation of NBs during mitosis [101, 102].

Given the complexity of physicochemical conditions, the manipulation of PTMs is an intriguing approach to influence LLPS. Thus, it is pivotal to understand the possible mechanisms in cancer-related PTMs associated with LLPS (Table 2).

Cancer cells can undergo unrestricted division [20, 133‐136], which can occur through gene mutations that activate oncogenic receptor tyrosine kinases (RTKs) and the downstream MAPK signaling involving RAS proteins.

Adaptor proteins involved in RTK and RAS signaling, such as LAT, GRB2 and SOS, undergo phase separation during RTK activation [137]. This phenomenon increases the interaction time between SOS and RTK/RAS, providing a mechanism for kinetic proofreading during RTK activation [125, 138] and preventing the spontaneous membrane localization of SOS, and the downstream activation of RAS. Interestingly, carcinogenic RTK mutations resulting from chromosomal rearrangements cause the loss of membrane localization but not its ability to stimulate downstream pathways. Mechanically, these condensates can assemble the RAS-activating complex GRB2/SOS1, which activates the RAS-MAPK signaling in a membrane-independent manner [123]. Moreover, RTK fusion oncoprotein granules enable the activation of RTK signaling [123, 139]. The close binding to RTK oncoprotein condensates allows GRB2 to concentrate key downstream molecules, achieving the constitutive activation of RAS-MAPK signaling in cancer cells (Fig. 4A). Therefore, BMCs provide a new method for modulating cancer-promoting signaling in a spatially restricted manner.

The ability to invade and metastasize allows the tumors to develop distantly, and the epithelial–mesenchymal transition (EMT) programs are commonly involved [140]. Activated by EMT, the transcription coactivators YAP and TAZ facilitate metastasis [141, 142]. Hu et al. found that YAP fusion proteins undergo LLPS in the nucleus and that the IDR provided by the partner of YAP is required for assembly. This aggregation promotes the YAP/TAZ-specific transcriptions and attenuates metastasis [68]. Similarly, another study revealed that the phase separation of DDX21 activates MCM5, thus triggering EMT signaling and modulating the colon cancer metastasis (Fig. 4B)[143]. Besides, SGs are also involved in malignant invasion and metastasis. EMT markers Cadherin, Vimentin, Snail and Slug are suppressed under SG core component G3BP1 depletion, implying the role of G3BP1 in tumor metastasis [144]. Moreover, G3BP modulates mRNA stability under stress conditions and facilitates the invasion of cancer cells [145]. These carcinogenic mechanisms provide new explanations for tumor metastasis, as well as the inspiring ideas for models of cancer progression regulation by the BMCs.

Cancer cells not only promote their growth but also modify tumor-suppression pathways [20]. By inhibiting tumor suppressors such as SPOP, p53, and RB1 [146‐148], cancer cells escape intrinsic growth limitations. P53, one of the most well-known tumor suppressors, inhibits tumorigenesis via transcriptional activation, which leads to the disorders of apoptosis, cell cycle, and cell senescence. Tumor-associated stress significantly triggers p53 aggregation [149‐154]. These findings demonstrate that the disruption of particular BMCs may cause cancer (Fig. 4C). Further studies are needed to validate this approach with other tumor suppressors and to test its potential applications.

Cellular senescence is considered an anticancer mechanism that maintain homeostasis and is associated with cell cycle arrest. The initiation and maintenance of cellular senescence rely on the frequent damage to the P53/Rb signaling pathway. Increased accumulation of 53BP1 in the nuclear foci after DNA damage can activate p53 and has recently been shown to regulate the cellular senescence via LLPS (Fig. 4D) [155].

Cancer cells can overcome the cell senescence and death via telomerase or alternative methods for lengthening telomeres (ALT) [156‐158]. Multivalent interactions between SUMO and SUMO-interacting motifs were observed in the formation of ALT-associated PML bodies on telomeres in cancer stem cells (Fig. 4E) [159]. The fusion of PML bodies enables the clustering of telomere elements and the recruitment of DNA helicases, and other molecular machinery to extend the length of telomeres [160]. This finding suggests that cancer stem cells achieve replicative immortality through the unchecked cell division, and that this process is associated with LLPS.

Common epigenetic modifications include histone modifications, DNA methylation, and RNA interference [161, 162]. Interactions between epigenetic modifications and their corresponding reader proteins also exhibit polyvalent interactions. M⁶A, known as the most common mRNA modification [163], alters the mRNA structure and interacts with multiple other mRNA modifications and proteins. This modification facilitates YTHDF protein phase separation, further contributing to the forming of various RNA–protein granules, including P bodies and SGs [74, 164]. In addition, YTHDC1 can undergo LLPS in the nucleus by interacting with m⁶A-modified mRNAs. This interaction results in the formation of nuclear YTHDC1-m⁶A condensates (nYACs), which are significantly enhanced in acute myeloid leukemia (AML) cells (Fig. 4F) [105].

The immune system employs the RLR-MAVS and cGAS-STING signaling pathways for protection against microbial invasion and support tumor immune surveillance [165‐167]. However, tumors often escape immune clearance surveillance. Recent findings by Meng et al. revealed that neurofibromin 2 (NF2) facilitates innate immunity by eliminating tank-binding kinase 1 (TBK1) activation. It is the missense mutations in the FERM domain of NF2 (NF2m) that robustly inhibit the STING-initiated antitumor immunity via the NF2m-IRF3 condensates formations (Fig. 4G), suppressing the TBK1 activation [130]. This offers novel insights into NF2-related cancer treatments.

Notably, inflammation often plays a dual role in cancer. Overproduced in various inflammatory tissues, the reactive oxygen species (ROS) may accelerate the genetic mutations of cells, making them more aggressive and malignant [168]. However, recent research indicates that the PML NB may function as a sensor for ROS in two ways: protecting cancer cells from excessive ROS or promoting ROS-induced apoptosis (Fig. 4H). Given the lack of in-depth research in this field, further tumor microenvironment exploration is required to understand these processes fully.

Tumor-associated viruses, such as human papillomavirus, Kaposi sarcoma herpesvirus, and Epstein–Barr virus (EBV), influence tumor progression through LLPS [169‐171]. In EBV proteins such as EBNA2 and EBNALP, LLPS regulates host gene expression, forming biomolecule condensates at Runx3 and MYC SE sites to regulate viral and cellular gene transcription (Fig. 4I). Further, the LLPS of EBNA2 can influence the alternative splicing of the pre-MPPE1 gene in cancer [170].

Vascularization, also known as angiogenesis, is crucial for supplying tumors with nutrients and oxygen for growth. Vascular endothelial growth factor (VEGF) is the leading factor responsible for rapid nutrient supply. Mounting evidence has indicated a correlation between BMC formation and angiogenesis. For example, the constitutive expression of the transcription factor (TF) MYC in metastasizing cells can lead to VEGF transcription by potentially forming phase-separated transcription condensates, promoting promotes angiogenesis [172]. Similarly, the use of 1,6-hexanediol, an inhibitor of LLPS, has recently been shown to regulate angiogenesis by inhibiting cyclin A1-related endothelial functions as well as condensates with BRD4, indicating that targeting condensates can block critical reactions (Fig. 4J) [173, 174].

Genomic instability contributes to tumor progression. Genomic translocations and rearrangements can lead to the fusion between the IDR of one protein and the DNA- or chromatin-binding domain of another [175]. This fusion acts as a TF, eliciting LLPS and attracting additional partners to initiate transcriptional programs that ultimately contribute to tumorigenesis. A typical example is the NUP98 fusion oncoprotein (FO), which occurs in 50% of patients with chemotherapy-resistant AML [176‐179]. FOs demonstrate that malignancies establish cancerous TF condensates [83, 180, 181] and attenuate aberrant chromatin organization (Fig. 4K).

Cancer cells can escape apoptosis by forming SGs (a form of MLOs) when exposed to extreme conditions, such as high temperatures, toxins, mechanical damage, or other stresses. For example, the Y-box binding protein 1 (YB-1) interacts with the 5'-untranslated region (UTR) of G3BP1[182], leading to the increased expression of G3BP1 and SGs, which is elevated in human sarcomas [183‐185]. Consequently, these cancer cells survive hyperproliferation, chemotherapy and other various stressful conditions. Additional studies on prostate cancer have demonstrated that the m⁶A-modified androgen receptor (AR) mRNA phase separated with YTHDF3, while the unmodified AR mRNA phase separated with G3BP1 to survive AR pathway inhibition stress (Fig. 4L)[186]. Collectively, SGs may serve as novel targets for cancer biology investigations.

Malignant cells undergo metabolic reprogramming [187], thereby attracting considerable interest in tumor-related research in the past decades [188]. For example, the reduction of glutaminase-1 (GLS1) enables cancer cells to survive under prolonged glutamine deprivation stress [189, 190]. Wang et al. reported that the lncRNA GIRGL promotes the LLPS of GIRGL-CAPRIN1-GLS1 mRNA to suppress GLS1 translation, thus adapting to an adverse glutamine-deficient environment (Fig. 4M)[191]. CAPRIN1, an RNA-binding protein involved in the SG formation via LLPS, plays a role in this metabolic adaptation. Therefore, alteration of cell adaptation to an adverse metabolic environment is possible by targeting condensates.

Considering that various regulatory mechanisms of LLPS are closely associated with tumorigenesis, it is imperative to explore therapeutic approaches against abnormal LLPS. These strategies can be categorized into three main approaches (Table 4).

Although LLPS and traditional vesicles are two different concepts with distinctive definitions, the vesicular trafficking role of LLPS is still rarely described and attractive. Conventional approaches typically utilize nanoscale carriers that are confined within the compartments of the intranuclear body. Nevertheless, recent findings have demonstrated that micron-scale polypeptide clusters, formed through phase separation, possess the ability to traverse the cell membrane via a non-canonical endocytic pathway. These clusters undergo glutathione-induced release of their cargo and exhibit the capacity to rapidly incorporate various macromolecules into microdroplets, such as RNA, small peptides and enzymes [223]. Loaded with polysomes, they can provide new approaches for vaccine carriers based on mRNAs and intracellular transportations for cancer treatments.

Likewise, as previously mentioned, droplets of drugs formed by LLPS can unexpectedly raise the inner drug concentration up to 600 times higher than that outside the condensate [202]. Furthermore, MED1 predominantly acts on oncogene promoters, thereby enabling cisplatin to ultimately target the corresponding DNA through its toxic platinum atoms, effectively assaulting the vital components of the cancer cells. Besides, the phosphopeptide KYp has been observed to induce LLPS level at the cell membrane, thus enhancing the permeation and internalization of the peptide drug [224]. KYp has the ability to interact with alkaline phosphatase, resulting in the dephosphorylation and in situ self-assembly at the cell membrane [224]. The process induces the aggregation of alkaline phosphatase and the separation of proteolipid phases at the membrane, ultimately enhancing membrane leakage and facilitating the entry of the peptide drug. These great discoveries provide inspirations for designing drug delivery systems and more similar ideas are worth exploring.