Intrinsically Disordered Regions of Zinc Finger Protein 687 in Phase Separation: A Potentially Pathogenic Driver of Paget’s Disease of Bone

Abstract

Paget’s disease of bone (PDB) is a chronic skeletal disorder characterized by abnormal bone remodeling and structural deformities, yet its underlying molecular mechanisms remain poorly understood. Genetic mutations in regulatory genes have been implicated in the PDB, potentially disrupting osteoclast activity and leading to excessive bone resorption and abnormal bone formation. Current research has largely focused on clinical phenotypes and downstream pathways, relying on molecular biology approaches and omics analyses to elucidate mechanisms. However, the impact of genetic mutations on the biophysical properties and functional dynamics of biomacromolecules remain unclear. Phase separation (PS), a process essential for biomolecular condensate organization and cellular function, has been implicated in various diseases but is unexplored in the PDB. This study investigates the PS properties of ZNF687, a zinc finger protein associated with PDB. Structural predictions and sequence analyses identified intrinsically disordered regions (IDRs) within ZNF687 that may drive PS, while protein interaction predictions suggest its involvement in bone remodeling pathways. These findings provide a novel perspective linking PS to PDB pathogenesis and highlight ZNF687 as a potential target for further investigation.

Paget’s disease of bone (PDB) is the second most prevalent metabolic bone disorder worldwide with a higher prevalence in certain geographic regions such as Europe, North America, and Australia. , PDB is estimated to affect approximately 1.5–8.3% of individuals over the age of 55 globally and affects men more frequently than women. The disease is characterized by excessive and disorganized bone remodeling, clinically manifested as bone pain; deformities; fractures; and in severe cases, neurological complications due to bone expansion and nerve compression. While the exact etiology of PDB remains unclear, genetic and environmental factors are believed to play key roles in its development. Mutations in genes such as SQSTM1, TNFRSF11A, and OPTN have been implicated in familial and sporadic forms of the disease, − disrupting normal osteoclast function and leading to pathological bone resorption and formation. As previously reported, the underlying mechanisms of PDB are wide-ranging and poorly understood, and patients without identifiable genetic mutations are often diagnosed as having idiopathic PDB. Additionally, environmental factors such as viral infections and occupational exposures have been proposed as contributing to the disease, but these hypotheses remain controversial and unproven. Complications such as secondary osteoarthritis, hearing loss, and an increased risk of bone sarcoma further add to the disease burden. PDB has gradually emerged as a significant public health challenge, particularly in aging populations, and represents a critical obstacle for clinicians and researchers aiming to prevent disease progression and improve patient outcomes.

Many pathogenic factors and key regulatory mechanisms in Paget’s disease of bone remain poorly understood, and there is a lack of precise and effective diagnostic and therapeutic approaches. Current research primarily focuses on clinical phenotypes and downstream signaling pathways, often elucidating the underlying mechanisms through genetic studies, animal models, and molecular biology experiments. These studies have provided insights into genetic mutations and altered bone remodeling processes associated with the disease. However, how mutations themselves affect the function of biomacromolecules remains mostly unclear. Intracellular organization is crucial for maintaining cellular homeostasis and supporting dynamic biological processes, especially for polymer chemistry of bone for bone regeneration. One emerging mechanism underpinning this organization is liquid–liquid phase separation (LLPS), which facilitates the formation of biomolecular condensates, including stress granules, nucleoli, and transcriptional hubs. Unlike membrane-bound organelles, these condensates form through reversible interactions between proteins and nucleic acids, enabling the compartmentalization of cellular processes without physical barriers. LLPS has been implicated in diverse cellular functions such as signal transduction, RNA metabolism, and chromatin organization. , Dysregulation of phase separation has been implicated in several diseases, including neurodegeneration, cancer, and bone disorders. −

ZNF687, a zinc finger protein, has garnered attention for its role in transcriptional regulation and its ability to influence cellular pathways through phase separation. , Zinc finger proteins are known for their DNA-binding capabilities and often function as transcription factors. We used predicted local distance difference test (pLDDT) (Figure A), predicted alignment error (PAE) plots (Figure B), and in silico AlphaMissense predictions (Figure C) to provide a comprehensive understanding of the structural stability and functional flexibility of ZNF687. The model reveals a central, rigid core likely composed of tightly packed structural elements surrounded by more flexible regions. Many zinc finger proteins contain intrinsically disordered regions (IDRs), which are key facilitators of phase separation due to their flexible, multivalent interactions. − These IDRs can undergo phase separation, forming biomolecular condensates that compartmentalize specific cellular reactions without physical barriers. Therefore, ZNF687, which consists of IDR activation domains, can regulate fundamental mechanisms to regulate gene expression through the LLPS capacity of their IDR-containing activation domains.

1.

Structure of ZNF687 predicted by AlphaFold2. (A) Predicted structure of human ZNF687 protein from the AlphaFold2 database. (B) The predicted alignment error (PAE) plots for the predicted ZNF687 models by AlphaFold2 are shown for the human ZNF687 protein. (C) In silico AlphaMissense predictions of ZNF687 pathogenicity probabilities for all possible single-amino acid variants are shown as a heat map with a color code range from red to deep blue.

Recent studies have highlighted the involvement of ZNF687 in Paget’s disease of bone (PDB), a chronic disorder characterized by abnormal bone remodeling and metabolism. − Mutations in ZNF687 have been linked to severe, polyostotic forms of PDB, often accompanied by an increased risk of developing giant cell tumors (GCTs). Studies in vitro have shown that mutant osteoclasts harboring ZNF687 mutations can develop up to 150 nuclei, a distinct feature in affected PDB patients. The pathogenic effects of ZNF687 are believed to arise from its role in dysregulated bone remodeling, potentially disrupting normal cellular signaling pathways involved in osteoclast differentiation and bone turnover. Given the presence of intrinsically disordered regions (IDRs) in ZNF687, the protein is hypothesized to undergo LLPS, contributing to its functional roles in gene regulation and disease mechanisms. Understanding whether ZNF687 participates in such processes through phase separation could reveal novel insights into its biological functions and disease implications.

Hypothesis

Following clues from existing studies, the current hypothesis suggests that phase separation dysregulation may represent a key pathogenic mechanism in Paget’s disease of bone. In this model, genetic mutations and microenvironmental changes disrupt the self-condensation or assembly of biomolecules, particularly within the intrinsically disordered regions (IDRs) of ZNF687. These disruptions interfere with normal phase separation processes. Targeted modulation of phase separation in ZNF687 could offer a promising therapeutic strategy for managing Paget’s disease of bone.

Evaluation of the Hypothesis

Structural Predictions of the Human ZNF687 Protein Suggest That It Has the Potential to Undergo Phase Separation

The structure of human ZNF687 reveals a central, rigid core likely composed of tightly packed structural elements surrounded by more flexible regions (Figure ). In Figure A, red regions represent stable, ordered structural elements located in the protein’s core, suggesting that these regions contribute to the overall structural stability. These elements are likely composed of structured α-helices, β-sheets, or other compact motifs. In contrast, the blue regions on the protein’s surface represent more flexible, loosely structured areas. These regions, which may be disordered or dynamic, could facilitate interactions with other proteins or molecules, highlighting the protein’s functional versatility. The predicted local distance difference test (pLDDT) reflects the reliability of the predicted structure with higher values indicating greater confidence in the prediction. Regions with pLDDT scores above 90 correspond to the more confidently predicted, stable red regions, while areas with pLDDT scores below 50 indicate lower confidence and are likely to represent disordered or flexible regions (blue). These lower-confidence regions, often located on the periphery, may play a role in the protein’s functional plasticity and interactions within the cellular context. Additionally, the predicted alignment error (PAE) plots for the human ZNF687 protein models generated by AlphaFold2 are shown in Figure B. The PAE plot provides a measure of the predicted accuracy of the structural alignments, with each point representing the predicted error between two residues. Lower PAE values indicate higher confidence in the predicted spatial relationships between residues, while higher values suggest greater uncertainty in these alignments. In the PAE plot for ZNF687, the lowest PAE values are observed in the 500–600, 800, and 1000 residue ranges in Figure B, suggesting high confidence in the predicted structural alignment in these regions. These areas correspond to stable, well-defined portions of the protein, which align with the red, high-confidence regions observed in Figure A. The low PAE values in these regions reflect a high degree of structural certainty, indicating that the protein’s central core and certain functional domains are well-predicted and likely to be rigid and stable. These regions may play critical roles in the protein’s DNA-binding activity or interactions with other cellular components. In contrast, regions outside these ranges with higher PAE values are predicted to have greater structural uncertainty, indicating more flexible or disordered regions. Such flexibility may be functionally relevant, allowing the protein to undergo conformational changes or interact with multiple partners. The AlphaMissense tool was utilized to predict the potential impact of single-amino acid substitutions on the function of the ZNF687 protein, with the resulting pathogenicity probabilities represented by a heatmap in Figure C. The color intensity in the heatmap ranges from red to deep blue, reflecting the predicted pathogenicity of each variant. Red regions correspond to residues with the highest predicted pathogenicity, indicating that mutations in these positions are likely to disrupt the protein’s function. Notably, the highest pathogenicity scores are observed in the 500–600, 800, and 1000 residue ranges, suggesting that variants in these regions may significantly affect the protein’s stability, structure, or its interaction with DNA and other molecules. In contrast, deep blue regions represent residues predicted to have minimal impact on the protein’s function, suggesting that substitutions in these areas are less likely to be pathogenic. Overall, consistent findings suggest that the flexible regions are intrinsically disordered regions of ZNF687. IDRs often contain multiple interaction sites, such as repetitive motifs (e.g., low-complexity regions or prion-like sequences), which enable them to interact with other proteins, nucleic acids, or cofactors in a cooperative manner. This multivalency facilitates the formation of complex networks of weak transient interactions that are essential for phase separation.

The Phase Separation Propensity of ZNF687 Has Been Consistently Identified Across Multiple Prediction Tools

IDRs tend to be rich in polar, charged, and hydrophobic residues. These properties promote liquid–liquid phase separation by driving weak, reversible interactions between molecules. To identify specific mutation-prone regions based on amino acid composition, human ZNF687 protein was analyzed using Composition Profiler, comparing it to the SwissProt (blue) and DisProt 3.4 databases (red) (Figure A). The plot highlights the relative enrichment and depletion of individual amino acid residues, with positive values indicating enrichment (>0) and negative values indicating depletion (<0) compared to the reference data sets. The residues are arranged in order of increasing disorder-promoting potential, offering valuable insights into the structural characteristics and potential functional regions of ZNF687. The analysis reveals a notable enrichment of cysteine (C), histidine (H), and proline (P) residues. This enrichment is significant as cysteine and histidine are essential for the formation of zinc finger domains, which are critical for ZNF687’s DNA-binding function. Proline, often found in loops and turns, may contribute to structural stability. In contrast, the analysis shows a depletion of tryptophan (W), tyrosine (Y), and isoleucine (I) residues, which are typically associated with hydrophobic or functional regions. The depletion of these residues may suggest reduced interaction with hydrophobic environments or fewer aromatic interactions within the protein.

2.

In silico analysis of the human ZNF687 sequence. (A) The amino acid composition of ZNF687 relative to the SwissProt and Disprot 3.4 (data set with disordered proteins) databases was plotted using Composition Profiler. Enrichment (>0) and depletion (<0) of given amino acid residues relative to the database is also shown; the amino acid residues are arranged in increasing order of disorder-promoting potential. (B) Uversky plot of the mean charge versus hydropathy scores; the absolute net charge of 105 completely ordered proteins was represented as blue circles, and 54 completely disordered proteins were represented as purple circles (The black solid line represents the border between the ordered and disordered proteins). The red dot corresponds to ZNF687. (C) The intrinsic disorder profile of ZNF687 was generated by the PONDRpredictor using the VL-XT algorithm. The value above 0.5 indicates a high probability of disorder. (D) Prediction of the phase separation propensity of the ZNF687 sequence. Graph illustrating PScore (red), catGRANULE (purple), and FuzDrop (blue) analysis results for the ZNF687 amino acid sequence. Values above 0 (PScore in red) and above 4 (catGRANULE in purple) indicate a putative tendency of protein fragments for driving LLPS, and values (pDPs in blue) above 0.6 indicate a probable tendency. (E) Summary of phase separation predicted total propensity score of ZNF687 generated by catGRANULE, PScore, FuzDrop, and PSPredictor. (F) Phase separation potential of ZNF687 generated by PhaSePred. (G) Summary of self- and part-phase separation prediction generated by PhaSePred in the eight-feature and 10-feature predictions. (H) Summary of score and rank prediction generated by PhaSePred containing nine predictors. (J) Major reported PDB-associated mutations (S242I, R331W, P665L, Q784E, P937R, , and R939C) are located within regions predicted to have LLPS or prion-like properties, as identified by catGRANULE and PLAAC, respectively. Mutations shown in bold indicate a prediction value greater than zero, while those in regular font represent prediction values equal to or below zero. In the catGRANULE plot, the rectangular region denotes LLPS propensity, whereas in the PLAAC plot, peak regions indicate prion-like domain propensity.

A Uversky plot in Figure B is a graphical representation used to categorize proteins as either folded (globular) or intrinsically disordered based on their calculated mean net charge and mean hydropathy. The plot includes data from 105 completely ordered proteins (blue circles) and 54 completely disordered proteins (purple circles). The black solid line represents the boundary between ordered and disordered proteins, distinguishing regions where proteins are predicted to adopt either a stable or a flexible conformation. The ZNF687 protein is represented by a red dot, positioned near the ordered region but close to the boundary line. This placement suggests that ZNF687 exhibits characteristics of a predominantly ordered protein with possible regions of structural flexibility. The proximity to the boundary indicates that while the protein maintains an overall stable structure, specific segments may adopt dynamic conformations, which could facilitate functional interactions or regulatory mechanisms. To evaluate the intrinsic disorder profile of the ZNF687 protein, we employed the PONDR predictor using the VL-XT algorithm. The disorder probabilities are plotted for each residue in Figure C, with scores above 0.5 indicating a high likelihood of intrinsic disorder. The analysis reveals prominent peaks of disorder in the residue ranges 200–400, 500, 900, and 1100–1200, indicating highly flexible regions that lack a defined secondary structure. These disordered segments are likely involved in facilitating dynamic interactions with other proteins or DNA, enhancing the functional adaptability of ZNF687.

Phase separation propensity refers to the likelihood that a protein will engage in LLPS within a cellular environment. This phenomenon is primarily driven by molecular interactions, particularly those involving intrinsically disordered regions and specific amino acid sequences that facilitate multivalent binding, either with the protein itself or with other molecules. In Figure D, the phase separation propensity of the ZNF687 sequence was predicted by using three different methods: PScore (red), catGRANULE (purple), and FuzDrop (blue). − The graph illustrates the analysis results for the ZNF687 amino acid sequence, with values above 0 for PScore (red) and above 4 for catGRANULE (purple) indicating a putative tendency of protein fragments to drive LLPS. For FuzDrop (blue), values above 0.6 suggest a probable tendency for phase separation. The PScore analysis reveals several regions, particularly residues 500 and 900, with values above 0, indicating a potential phase separation activity. Similarly, catGRANULE and FuzDrop highlight specific segments, particularly in the 200–400, 500, 900, and 1100–1200 residue ranges, suggesting that these regions of ZNF687 have high phase separation propensities. These regions may play a critical role in the protein’s ability to participate in LLPS, which could be crucial for its cellular localization and function. In Figure E, a summary of the phase separation predicted total propensity scores for ZNF687 is provided, as calculated by the catGRANULE (1.196), PScore (3.36), FuzDrop (0.9986), and PSPredictor (0.8135) tools.

The PhaSePred software was further used to assess the phase separation potential of ZNF687 by integrating scores from several phase separation-related prediction tools, including catGRANULE, PLAAC, PScore, ESpritz (Disprot), Hydropathy, SEG, Charged Residue, Phosfrequency, and DeepPhase, as shown in Figure F. The results confirmed the predicted phase separation propensity of ZNF687, suggesting the potential for the formation of dynamic, phase-separated structures. Notably, high predicted phase separation scores in the catGRANULE, PLAAC, PScore, and ESpritz (Disprot) data sets indicated a strong tendency for ZNF687 to drive LLPS. In contrast, lower scores in Hydropathy, Charged Residue, and DeepPhase analyses suggest that ZNF687 has relatively low hydropathy and fewer charged residues, which may influence its phase separation characteristics, as seen in Figure H. Figure G presents the self- and part-assembling phase separation predictions for ZNF687, generated by PhaSePred using both the eight-feature and 10-feature models. The analysis revealed distinct patterns in phase separation propensity between the two models. The 10-feature model (PS-self) yielded a high score for self-phase separation, indicating that the entire ZNF687 protein has a strong potential to undergo LLPS. In contrast, the eight-feature model (PS-part) demonstrated a higher score for part-phase separation, suggesting that specific regions of ZNF687 are more prone to phase separation, even if the full protein does not exhibit the same propensity. Moreover, major reported PDB-associated mutations (S242I, R331W, P665L, Q784E, P937R, , and R939C) are located within regions predicted to have LLPS or prion-like properties, as identified by catGRANULE and PLAAC, respectively (Figure I). The consistency of these findings across multiple prediction methods further supports the hypothesis that ZNF687 contains functional domains capable of contributing to LLPS, highlighting its potential to form dynamic, phase-separated structures that may be critical for its cellular functions.

Nuclear Aggregates Serve As a Potential Marker for the Phase Separation Propensity of ZNF687

Immunofluorescence images of ZNF687 were acquired in the MCF7 and U-2 OS cell lines as part of the Human Protein Atlas in Figure A. Cells were costained with DAPI to visualize nuclear DNA, ZNF687 to label the protein, and microtubules to identify the cytoskeleton. The results show that ZNF687 predominantly aggregates within the nucleus, with a significantly higher intensity observed in the nuclear compartment compared to the cytoplasm. These nuclear aggregates appear as distinct condensates, supporting the hypothesis that ZNF687 may be involved in liquid–liquid phase separation (LLPS). The stronger fluorescence intensity in the nucleus suggests that ZNF687 forms more stable or larger condensates within the nucleus, potentially influencing its nuclear functions. In contrast, the cytoplasm shows weaker fluorescence, indicating that the ZNF687 aggregates may be more dynamic or less concentrated in this region. Interestingly, ZNF687 exhibits distinct expression patterns between the two cell lines, MCF7 and U-2 OS, suggesting that the protein’s aggregation and localization may vary depending on the cellular context. Furthermore, a previously published study also demonstrates a punctate, condensate-like distribution of ZNF687 in osteoclasts from a PDB patient with a coexisting giant cell tumor (Figure B). This differential expression implies that ZNF687 may have cell line-specific functions or regulatory mechanisms, further highlighting its potential involvement in context-dependent cellular processes. These findings underscore the specificity of ZNF687 in different cellular environments and support its role in forming phase-separated nuclear structures, which may be critical for its functional dynamics.

3.

Examples of human ZNF687 protein showed clear condensates. (A) The images were acquired in the MCF7 and U-2 OS cell line as part of the Human Protein Atlas. The immunofluorescence profiles suggest that ZNF687 is likely to form condensates in the nucleus. (B) Immunostaining shows the distribution of ZNF687 in osteoclasts from a PDB patient with a coexisting giant cell tumor. Three regions were zoomed as R1, R2, and R3. Adapted with permission from ref .

Numerous Molecules Associated with ZNF687 Exhibit Phase Separation Potential in the Context of Paget’s Disease of Bone

A comprehensive interaction network of ZNF687 was constructed using the Search Tool for the Retrieval of Interacting Genes and Proteins (STRING), revealing over 30 interacting proteins (Figure A). Among these, several key interactors are associated with the role of ZNF687 in Paget’s disease of bone. Notably, KDM5C, a histone demethylase involved in chromatin remodeling and transcriptional regulation, interacts with ZNF687, suggesting its role in epigenetic modulation. Similarly, TSPYL1 and TSPYL2, which are implicated in transcriptional regulation and chromatin modification, associated with ZNF687, potentially influencing cellular differentiation and stress responses. The interaction with CSNK2A, a casein kinase involved in signaling pathways, suggests that ZNF687 may play a role in regulating signal transduction in the global development delay. , Additionally, BRD3, a bromodomain-containing protein critical for transcriptional activation via chromatin remodeling, interacts with ZNF687, supporting its involvement in transcriptional coregulation. ZMYND8, known as a transcriptional corepressor, partners with ZNF687 in suppressing specific gene expression programs, further emphasizing its regulatory functions through its association with the P-TEFb complex. Lastly, ZNF592, a zinc finger protein associated with transcriptional regulation, underscores potential role of ZNF687 in modulating transcriptional networks of genes expression in a complex developmental pathway. Together, these interactions reveal participation of ZNF687 in essential biological processes, including chromatin remodeling, transcriptional regulation, and signal transduction, offering new insights into its functional significance in Paget’s disease of bone.

4.

Phase separation prediction of ZNF687 interactors. (A) Comprehensive ZNF687 interaction network. (B) Phase separation potential prediction of all 20 ZNF687 interacting proteins generated by PhaSePred. (C) The immunofluorescence images of all 20 ZNF687 interactors were acquired in the U-2 OS cell line as part of the Human Protein Atlas.

To evaluate the propensity of ZNF687 interactors, which can undergo phase separation, we utilized PhaSePred, a computational tool that predicts phase separation potential based on sequence features such as amino acid composition and intrinsically disordered regions. The analysis identified several ZNF687 interactors with high phase separation scores, suggesting their potential involvement in biomolecular condensate formation (Figure B). Proteins with intrinsically disordered regions and multivalent interaction motifs were particularly enriched among the interactors with high scores, aligning with the known characteristics of phase-separating proteins. Key interactors with strong phase separation potential include members of the BRD family (e.g., BRD3 and BRD4), known for their roles in chromatin organization and transcriptional regulation, and MBD2, which participates in epigenetic processes. Additionally, proteins such as TSPYL2, CHD3, BRPF3, and HDAC1 were all associated with chromatin remodeling and transcriptional modulation in homeostatic cell health, independently exhibiting high risk for phase separation (Published proteins related to phase separation are summarized in Table ). Notably, ZMYND8, a transcriptional corepressor, was also identified as a high-scoring interactor, further supporting its potential to mediate phase-separated compartments. The analysis revealed a common cluster of interactors, including BRD family proteins, MBD proteins, CSNK2A, TSPYL proteins, and GATA proteins, highlighting functional convergence in transcriptional and chromatin-related pathways. These findings suggest that ZNF687 and its high-scoring interactors may play critical roles in the dynamic organization of nuclear compartments through phase separation, which is mostly found in Figure A.

1. Summary of Published Proteins Related to Phase Separation.

biomolecular condensates	building block for LLPS	complex involved	model for verification of phase behavior	results achieved	LLPS in disease	ref
BRD1	IDRs within BRD1	the HBO1 histone acetyltransferase complex, which are BRD1, EAF6, ING4 and HBO1	- colocalization in MKN45 or 293T cells	- rapid fluorescence recovery of EGFP-BRD1 puncta, confirming identity as liquid condensates	- BRD1 LLPS promotes chemoresistance in gastric cancer through OXPHOS genes upregulation.
			- fluorescence recovers after photobleaching to confirm liquid-like dynamics of condensates in vitro	- NIT2 overexpression reduces BRD1 puncta formation.
BRD4	IDR enriched in proline and glutamine content and acidic and basic stretches	SE condensates comprise BRD4 together with MED1, p300, CDK9/p-TEFb (Cyclin T1), and RNA Pol II	- colocalization of EGFP-BRD4 in HeLa cells	- 1,6-hexanediol causes loss of chromatin-bound BRD4	- pan cancer, especially breast cancer, where active oncogene transcription promotes proliferation	−
			- FRAP assays showing rapid fluorescence recovery of BRD4 foci		- atherosclerosis
			- droplet observation from purified BRD4-IDR–mEGFP fusion protein	- dynamic spherical droplets and reversible LLPS behavior in low-salt environments	- acute promyelocytic leukemia
MBD2	N-terminal glycine/arginine-rich (GR) IDR drives multivalent electrostatic interactions with MBD and coiled-coil (CC) domain	nucleosome remodeling and deacetylase (NuRD) complex, interacting with HDAC1, MTA2, and p66α	- FRAP in EGFP-MBD2 expresses NIH-3T3 cells	- liquid behavior of EGFP-MBD2 confirm by FRAP	- pan cancer (ovarian cancer), caused by excessive MDB2 condensate in sustaining a repressive chromatin state	,
			- droplet observation of MBD2 forming droplets in vitro	- purified MBD2 undergoes LLPS above ∼0.6 μM, with droplet size/number scaling with concentration
			- coarse-grained slab simulations to evaluate condensate formation and mapping key interdomain contacts in silico	- droplets dissolve in high-salt environments and reform upon low-salt environments
MBD3	C-terminal acidic (D/E-rich) IDR, MBD, and CC domain work together to enable multivalent electrostatic and hydrophobic interactions	integral subunit of NuRD complex, HDAC1, MTA2, and Mi-2 participate in NuRD	- colocalization of EGFP-MBD3 in NIH-3T3 cells and FRAP for liquid dynamics in cellular	- confirmation of liquid behavior of EGFP-MB3 puncta in the event of rapid FRAP recovery in live cells	not available
			- droplet observation of purified MBD3 protein in vitro	- purified MBD3 LLPS is concentration- and temperature-dependent with reversible droplets.
			- turbidity assays revealing dual UCST/LCST behavior in vitro
			- CG slab simulations showing stable condensates and mapping the acidic tail’s contacts with MBD/IDR regions in silico	- removal of the D/E tail enhances LLPS; deletion of the CC or MBD abolishes LLPS
DDX24	N-terminal domain (NTD), C-terminal domain (CTD), and IDR, which includes the E271 K mutation, provide multivalent, mainly electrostatic, interactions	DDX24 partitions into nucleolar condensates scaffolded by NPM1 oligomers and rRNA, driven by multivalent IDR–NPM1/rRNA interactions	- droplet formation in purified WT and E271 K DDX24 (±rRNA)	- FRAP shows gel-like internal dynamics in vitro	- mutiorgan venous
			- turbidity was accessed and fluorescent microscopy used to visualize droplets	- rRNA enhances droplet formation	- lymphatic defect (MOVLD) syndrome
			- rapid FRAP recovery, confirming liquid-like behavior	- DDX24 form droplets that dissolve in high salt and reform in low salt	- Budd-Chiari syndrome
			- live-cell FRAP on EGFP-DDX24 in HUVECs to compare dynamics of WT vs E271 K in nucleoli		- pathological endothelial behavior
HDAC1	IDRs, specifically the histidine-rich (H) and glutamate/aspartate-rich (E/D) clusters	Formation of YY1-HDAC1/3 complex that modulates YY1’s phase separation behavior at the METTL3 promotor in acute myeloid leukemia cells	- colocalization of HDAC1 with YY1 with or without HDAC inhibitor treatment	- HDAC1 depletion or CHI treatment causes excessive YY1 condensate formation due to poor YY1-HDAC1 binding.	- acute myeloid leukemia
			- granule formation assessment
ZMYND8	IDRs rich in aspartate/glutamate residues	ZMYND8 coassembles with acetylated NFkB/p65 at latent superenhancer loci and recruit the histone demethylase LSD1, forming a tripartite repressive condensate	- droplet observation in purified GFP-ZMYND8-IDR in low-salt environments	- reversible property of IDR droplets	- diet-induced obesity
			- FRAP evaluation of droplet upon washout of high salt or 1,6-hexanediol in vitro	- fast fusion and FRAP reveals molecular mobility and exchange dynamics	- colitis inflammation
			- live-cell FRAP of GFP-ZMYND8 in macrophages	- mutation of acidic residues to alanine converts liquid droplets into solid aggregates

The spatial distribution and organization of the 19 ZNF687-associated proteins localized to distinct cellular compartments, including the nucleus, cytoplasm, and cytoskeleton, with several displaying punctate patterns indicative of phase-separated condensates, as shown in Figure C. Notably, nuclear aggregates were prominently observed in proteins such as BRD1, BRD2, BRD3, and CHD3, which are linked to chromatin remodeling and transcriptional regulation, suggesting their potential roles in nuclear phase-separated hubs such as transcriptional condensates. Conversely, cytoplasmic localization with condensed patterns was observed in proteins such as DDX24, BRPF3, KDM5C, and MTA3, pointing to their possible involvement in cytoplasmic phase-separated structures, including stress granules and signaling complexes. Together, these findings provide compelling evidence that ZNF687 and its interactors may participate in phase-separated condensates, facilitating key processes such as transcriptional regulation, signal transduction, and cellular stress responses in the mechanism of Paget’s disease of bone.

Discussion

Phase separation (PS) of biomolecules, including proteins and nucleic acids, in cells has been shown to play a crucial role in various physiological processes. Aberrant phase separation often results in dysfunctional biomolecules contributing to pathogenesis. Our structural predictions of ZNF687, particularly via AlphaFold2, revealed a highly ordered core dominated by C2H2 zinc finger domains, known for their role in DNA-binding and protein–protein interactions. These domains mediate the protein’s involvement in chromatin remodeling and transcriptional regulation for the disease progression. The peripheral regions of ZNF687, which are predicted to be more flexible and disordered, could facilitate interaction of the protein with other cellular macromolecules and contribute to its functional adaptability. The analysis of predicted alignment error (PAE) supported this hypothesis, highlighting the well-structured nature of the core and the less defined, flexible regions that are likely to engage in multivalent interactions. These flexible regions may be particularly important for the protein’s involvement in dynamic cellular processes such as LLPS.

LLPS is a crucial mechanism underlying the formation of membraneless organelles, which regulate various cellular functions, including stress response, gene expression, RNA processing, and energy metabolism. Our analysis, utilizing phase separation prediction tools such as PScore and FuzDrop, revealed that ZNF687 exhibits a high propensity for phase separation, particularly in regions enriched with intrinsically disordered sequences, intricately linking to human health and disease. This predicted phase separation propensity aligns with a database demonstrating that ZNF687 forms nuclear condensates in MCF7 and U-2 OS cell lines. The nuclear aggregation of ZNF687 suggests a pivotal role in organizing nuclear processes, such as the regulation of transcription or chromatin remodeling, thereby influencing cellular states ranging from development to age-related disease. Notably, the context-dependent aggregation patterns observed across different cell lines underscore the tightly regulated nature of ZNF687’s phase-separating properties. This regulation is a hallmark of LLPS-driven mechanisms that govern cell survival and pathogenesis, potentially influenced by cellular redox states. Moreover, LLPS has emerged as a fundamental mode of cellular organization essential for processes like asymmetric cell division. Dysregulation of LLPS involving ZNF687 in these contexts may contribute to the disease pathology.

Our protein–protein interaction (PPI) analysis using STRING identified several key regulators of chromatin dynamics and transcription that interact with ZNF687, including BRD2/3, CHD3, BRPF3, TSPYL2, HDAC1, and ZMYND8. These interactions suggest that ZNF687 plays a role in modulating gene expression and cellular differentiation, orchestrating a complex network of transcription factors that enable cells to respond to both internal and external signals, maintain cell identity, and coordinate cellular activities. , Our study highlights the involvement of transcription factors in liquid–liquid phase separation (LLPS), where they contribute to the formation of transcriptional condensates. These condensates compartmentalize the transcriptional machinery, thereby facilitating efficient gene regulation. For instance, BRD3 has been shown to contribute to the formation of phase-separated structures in the nucleus, such as superenhancer hubs and chromatin condensates. Moreover, the transcriptional factors in this network are integral to the regulation of metabolism and oxidative stress, processes that are crucial for development , and metabolism disorder. Notably, HDAC1 has a well-established role in both metabolic pathways and oxidative stress responses, influencing cellular homeostasis and adaptation to stress.

The dysregulation of this complex transcriptional network has significant implications for diseases, particularly in the context of metabolic disorders such as Paget’s disease of bone (PDB). A key aspect of this network is the involvement of these transcription factors in the interaction of the processing of phase separation. Many of these factors regulate genes crucial for cell cycle control, apoptosis, and DNA repair. For example, BRD4 is a well-established example of a protein that undergoes phase separation. It forms transcriptional condensates at superenhancers and facilitates transcriptional regulation through liquid–liquid phase separation (LLPS). MBD2 and MBD3 are proteins involved in chromatin remodeling and nucleosome positioning. Evidence suggests their involvement in forming phase-separated chromatin condensates. Additionally, ZMYND8 is a protein known as an “epigenetic reader” that primarily functions by interacting with specific histone modifications, playing a crucial role in chromatin regulation by binding to modified histone tails and influencing gene expression; recent research indicates that ZMYND8 might also be involved in the formation of phase-separated chromatin domains, allowing for localized control of gene activity within the nucleus. Collectively, these findings emphasize the multifaceted role of ZNF687 in the pathogenesis of metabolic disorders, with its complex interactions with other transcription factors highlighting its potential involvement in regulating cellular processes essential for disease progression.

The analysis of mutation-prone regions of ZNF687, identified through AlphaMissense, further strengthens the hypothesis that alterations in the structure of ZNF687 could have pathogenic consequences. Mutations in the regions spanning residues 500–600, 800, and 1000–1100 may disrupt the protein’s structural integrity and impair its ability to engage in critical molecular interactions. Given the potential for such mutations to affect phase separation capacity of ZNF687, future studies should focus on elucidating how these mutations affect their biological functions and their possible connection to Paget’s disease of bone (PDB) through phase separation mechanisms. It will also be important to investigate how this transcription factor interacts with other proteins in the pathogenesis of PDB. Recent studies have highlighted that modulating biomolecular condensates represents a promising and innovative approach to drug discovery, with the potential to transform therapeutic strategies across a wide range of diseases. ,

Conclusions

Our study provides a comprehensive analysis of ZNF687, combining structural, phase separation, and protein interaction data to offer a deeper understanding of its biomacromolecule role in Paget’s disease of bone. ZNF687 appears to be a multifunctional protein with a central role in chromatin regulation and nuclear organization, potentially through phase separation. The interplay between its ordered and disordered regions suggests that ZNF687 may act as a molecular hub, coordinating the regulation of gene expression and other cellular processes. Future research should aim to validate these findings experimentally and explore the functional consequences of mutations in ZNF687, with the potential to uncover new therapeutic strategies for Paget’s disease of bone.

Glossary

Abbreviations

PDB

Paget’s disease of bone

LLPS

Liquid–liquid phase separation

PPI

Protein–protein interaction

PS

Phase separation

IDRs

Intrinsically disordered regions

pLDDT

Predicted local distance difference test

PAE

Predicted alignment error

GCTs

Giant cell tumors

STRING

Search Tool for the Retrieval of Interacting Genes/Proteins

#.

Jing Wang and Zheng Guo contributed equally.

No sources of funding were received for the preparation of this article.

The authors declare no competing financial interest.