Abstract Biomolecular condensates are membraneless compartments involved in a wide range of cellular processes. Despite their fundamental role in the spatiotemporal regulation of cellular functions, tools for precisely manipulating phase-separated condensates remain limited, and effective methods for discovering and functionalizing tunable phase separation modules from natural proteins are lacking. Here we present a rational engineering approach for androgen receptor (AR) and its clinically used drugs to create a chemical genetic platform, ARDrop, enabling condensates formation and dissolution. This platform is applied to a diverse set of proteins to achieve intended cellular functions, ensuring robust and long-lasting functionality through stable liquid-like properties. Our work develops a powerful toolkit for reversible manipulation of condensates that can be used for dissection of complicated cell signaling, laying the foundation for engineering designer condensates for synthetic biology applications. Subject terms: Protein design, Cell signalling, Synthetic biology, Intrinsically disordered proteins __________________________________________________________________ Tools for precisely manipulating phase-separated condensates remain limited. Here authors present an engineered system based on the androgen receptor, where clinically used drugs can be used to control the formation and dissolution of synthetic condensates, subsequently allowing control over various cellular processes. Introduction Changes in cell behaviors are governed by dynamic modulations in the activity of signaling and regulatory proteins^[42]1,[43]2. To properly study how these changes spread through molecular networks, we need tools that can manipulate these processes at spatiotemporal scales. Recently, biomolecular condensates that form distinct compartments to concentrate or exclude molecules have emerged as the fundamental mechanism regulating various biological processes^[44]3–[45]5. As such, the robustness and versatility of these condensates offer new opportunities for manipulating cellular states. A key feature of proteins forming biomolecular condensates is multivalency^[46]4,[47]5, achieved through multiple binding domains, oligomerization, or intrinsically disordered regions (IDRs) that facilitate weak interactions. Recently developed tools for the dynamic manipulation of condensates include light-controlled systems, such as those utilizing light-responsive Cry2/Cry2olig^[48]6–[49]8 or iLID-SspB^[50]9,[51]10 fused with intrinsically disordered regions. Chemical-induced systems based on the interaction between FRB and FKBP12 in the presence of rapamycin^[52]11–[53]13, BCL6 BTB domain and its ligands BI-3802 and BI-3812^[54]14,[55]15 have also been reported. While these tools are useful, each comes with specific limitations. Plant/fungal-derived systems offer orthogonality in mammalian cells but require two components or light activation and some show poor reversibility, whereas mammalian single-protein tools provide reversibility but lack orthogonality. Some tools transition from a liquid to gel or solid state over time, preventing the maintenance of a liquid-like functional state. Androgen receptor (AR) is a nuclear transcription factor that can translocate into nucleus and phase separate upon ligand(androgen) stimulation^[56]16–[57]19. Our previous work^[58]19 showed that the formation and dissolution of AR mutant condensates can be regulated by two generations of clinically used antiandrogens, but with low responsiveness to natural androgen ligands. This inspired us to design chemical-responsive synthetic AR condensates. First, we modified AR constructs to remove their own functions to avoid cross-reactivity within mammalian cells. By applying the principles uncovered in studies^[59]17,[60]19 that revealed the driving forces behind AR phase separation, we create fusions of minimal and tunable AR modules with functional domains, enabling us to control the formation and dissolution of condensates. This, in turn, modulates specific cellular functions, such as signaling transduction, stress granule formation and cell death in mammalian cells. Our work introduces ARDrop as a versatile and reversible chemical-genetic platform for investigating cellular states and provides a powerful tool for engineering biological systems. Results Design of chemogenetic tool ARDrop for inducible condensates Naturally occurring intrinsically disordered proteins (IDPs) constrain the programmability of phase behavior and may fail to produce orthogonal systems in vivo due to cross-reactivity between designed condensate components and functional elements of the cellular environment. To get past this barrier, we first sought to engineer constructs (Fig. [61]1A) that eliminates the original function of the AR protein while preserving its ability to undergo phase separation. Upon DHT (Dihydrotestosterone) stimulation, AR-WT translocated to the nucleus and forms distinct nuclear puncta (Fig. [62]1B). We deleted the nuclear localization signal to generate the AR-delNLS-mEGFP variant and observed the formation of large cytosolic AR condensates with DHT stimulation, due to the removal of constraints imposed by chromatin interactions (Fig. [63]1C, Supplementary Fig. [64]1). This observation parallels findings^[65]19 with the DNA-binding defective AR mutant (A574D)^[66]20, which also exhibited an enlarged puncta pattern (Fig. [67]1D, Supplementary Fig. [68]1). When combining NLS deletion with the DNA-binding defective mutation (AR-A574D-delNLS-mEGFP), or removing the entire DBD-hinge region (AR-delDBD-H-mEGFP), the AR variant retains its ability to undergo phase separation upon DHT stimulation (Fig. [69]1E, F). We compared nuclear AR (WT), AR(A574D), and constructed cytoplasmic condensates, finding that AR(A574D) and cytoplasmic puncta were larger than nuclear AR(WT) but retained liquid-like dynamics, with a shorter t[1/2] than the wildtype (Supplementary Fig. [70]1). To further demonstrate that the phase separation capability of AR condensates can be uncoupled from their function, we conducted immunofluorescence to assess active transcription markers in DHT stimulated condensates of AR wildtype and variants. In wildtype puncta, we observed significant enrichment of H3K27ac and RNA Pol II S5P; however, this enrichment was not detected in the condensates of cells expressing AR-A574D-mEGFP, AR-delNLS-mEGFP, AR-A574D-delNLS-mEGFP and AR-delDBD-H-mEGFP (Fig. [71]1G, H). Moreover, these transcriptionally inactive cytoplasmic puncta showed no colocalization with stress granules (Supplementary Fig. [72]2) or protein degradation sites (e.g., Hsp70, p62, or ubiquitin) (Supplementary Fig. [73]3). Then we use cytosolic AR constructs (AR-delDBD-H-mEGFP) as the starting point for subsequent optimizations. Fig. 1. Uncoupling AR phase separation and its original function. [74]Fig. 1 [75]Open in a new tab A Schematic representation of AR wildtype and engineered variants. B Time-lapse imaging of HEK293 cells stably expressing AR-WT-mEGFP (green), before and after DHT (dihydrotestosterone) stimulation (Up). Scale bar, 40 µm. Statistical analysis of the LLPS puncta number of AR-WT-mEGFP (Down), n = 30 fields (A total of 30 independent fields were acquired using the PerkinElmer Harmony™ high-content imaging system). C Time-lapse imaging of HEK293 cells stably expressing AR-delNLS-mEGFP (green), before and after DHT stimulation (Up). Scale bar, 40 µm. Statistical analysis of the LLPS puncta number of AR-delNLS-mEGFP (Down), n = 30 fields. D Time-lapse imaging of HEK293 cells stably expressing AR-A574D-mEGFP (green), before and after DHT stimulation (Up). Scale bar, 40 µm. Statistical analysis of the LLPS puncta number of AR-A574D-mEGFP (Down), n = 30 fields. E Time-lapse imaging of HEK293 cells stably expressing AR-A574D-delNLS-mEGFP (green), before and after DHT stimulation (Up). Scale bar, 40 µm. Statistical analysis of the LLPS puncta number of AR-A574D-delNLS-mEGFP (Down), n = 30 fields. F Time-lapse imaging of HEK293 cells stably expressing AR-delDBD-H-mEGFP (green), before and after DHT stimulation (Up). Scale bar, 40 µm. Statistical analysis of the LLPS puncta number of AR-delDBD-H-mEGFP (Down), n = 30 fields. G Immunofluorescence images of H3K27ac (red) in HEK293 cells stably expressed AR (WT/A574D/delNLS/A574D-delNLS/delDBD-H)-mEGFP (Left) (green). Scale bar, 5 µm. Images are representative of 3 independent experiments. Quantification of fluorescence intensity of AR (WT/A574D/delNLS/A574D-delNLS/delDBD-H)-mEGFP (green) and anti-H3K27ac (red) along the line indicated in merged image were shown (Right). Scale bar, 5 µm. H Immunofluorescence images of Pol II-S5P (red) in HEK293 cells stably expressed AR (WT/A574D/delNLS/A574D-delNLS/delDBD-H)-mEGFP (Left) (green). Scale bar, 5 µm. Images are representative of 3 independent experiments. Quantification of fluorescence intensity of AR (WT/A574D/delNLS/A574D-delNLS/delDBD-H)-mEGFP (green) and anti-Pol II-S5P (red) along the line indicated in merged image were shown (Right). Scale bar, 5 µm. Box plots display the median (center line), interquartile range (IQR; box boundaries between 25th and 75th percentiles), whiskers (1.5 × IQR from the box), and outliers (points beyond 3 × IQR). Source data are provided in [76]Source Data file. Following the decoupling of LLPS from AR function, we focused on simplifying the cytosolic AR constructs to their minimal LLPS components, enabling further flexible application across diverse biological systems. Previous studies^[77]19 show that full-length AR (920 aa) requires all domains (NTD, DBD and LBD) for efficient condensate formation, with the disordered NTD being particularly critical. Solution nuclear magnetic resonance (NMR)^[78]17 revealed transient intermolecular interactions mediated by aromatic residues especially in Tau-5 region of NTD and these aromatic tyrosine residues contribute to AR condensate formation. DHT binding triggers AR oligomerization^[79]17,[80]19 through N/C interactions^[81]21 (mediated by ^23FQNLF^27 motif) and LBD homodimerization, doubling AR valency without changing aromatic residue interactions—thus reducing the phase separation threshold. These results underscore the critical role of the ^23FQNLF^27 motif and Tau-5 in AR phase separation, which are necessary for LLPS formation (Fig. [82]2A). Previous results^[83]19 also showed that polyQ is not essential for AR condensate formation. When we deleted the sequence from polyQ to the region preceding Tau-1(44−102), phase-separated condensates retained according to experimental observations (Fig. [84]2B and Supplementary Fig. [85]4A). Guided by helical motifs identified through NMR^[86]17, we introduced four deletions (44−163, 226−347, 226−399 and 451−559) bypassing the helix region to assess their impact on AR phase separation (Fig. [87]2A). Deletions of 44−163 and 226−347 still exhibited significant puncta formation upon stimulation (Fig. [88]2B and Supplementary Fig. [89]4B). However, deletions of 226-399, 332-450 or 451-559 abolished phase separation (Fig. [90]2B and Supplementary Fig. [91]4A). Taken all these together, we finally get the simplified AR phase separation construct AR-del (44−163, 226−347, DBD-H), dubbed SynIDR, with a total length of 579AA (Fig. [92]2C). Fig. 2. Construction of chemical-modulated AR phase separation module. [93]Fig. 2 [94]Open in a new tab A Distribution of Short Helical Motifs and Aromatic and Tyrosine Residues within the Activation Domain of AR. Double-headed arrows indicate the regions deleted from the Activation Domain of AR (adapted from^[95]17). Tau5 fragment characterized by NMR spectroscopy is designated as Tau5*. B Statistical analysis of the LLPS puncta of different AR deletions at different time points under DHT treatment. Data are represented as mean ± SEM, n = 30 fields. C Time-lapse images before and after the addition of DHT to HEK293 cells stably expressing AR del (44−163, 226−347, DBD-H)-mEGFP (Up) (green). Scale bar, 40 µm. Statistical analysis of the LLPS puncta number of AR del (44−163, 226−347, DBD-H)-mEGFP (Down), n = 30 fields. D Annotation of W742L/C in the AR LBD (Up), and the pharmacological characteristics of androgen receptor driver mutations in response to antiandrogens and steroids (Down). E Time-lapse imaging of HEK293 cells stably expressing ARDrop (green), treated sequentially with 200 nM Bica (Bicalutamide) followed by 2 µM Enza (Enzalutamide). Scale bar, 40 µm. Statistical analysis of the LLPS puncta number of ARDrop (Right), n = 30 fields. F Time-lapse imaging of HEK293 cells stably expressing ARDrop (W742C) (green), treated sequentially with 200 nM Bica followed by 2 µM Enza. Scale bar, 40 µm. Statistical analysis of the LLPS puncta number of ARDrop (W742C) (Right), n = 276 fields. Box plots display the median (center line), interquartile range (IQR; box boundaries between 25th and 75th percentiles), whiskers (1.5 × IQR from the box), and outliers (points beyond 3 × IQR). Source data are provided in [96]Source Data file. Antagonists targeting the ligand-binding domain (LBD) of the androgen receptor (AR), commonly referred to as antiandrogens, represent key therapeutic strategy for prostate cancer treatment. However, clinically used antiandrogens inevitably lead to drug-resistant mutations within the LBD. AR(W742L/C) mutation was identified to confer an antagonist to agonist switch for bicalutamide^[97]22. As we previously reported^[98]19, Bicalutamide dose-dependently induce the formation of AR(W742L/C) condensates, which can be suppressed by second-generation antiandrogen Enzalutamide. Moreover, AR(W742L/C) is less sensitive to natural androgen (DHT) stimulation compared to AR(WT) in condensates formation^[99]22 (Fig. [100]2D). AR contains such compound-responsive modules, making it a versatile platform for modulating protein condensates. As expected, SynIDR with the chimeric W742L/C mutant LBD can undergo phase separation formation/dissolution in response to Bica/Enza. In HEK293 cells, untreated AR-SynIDR-W742L/C showed diffuse cytosolic localization and then rapidly formed significant condensates upon 200 nM bicalutamide treatment within 15 min (Fig. [101]2E, F). After 45 min of bicalutamide treatment, the addition of 2 µM enzalutamide caused rapid dissolution of the phase-separated droplets, with 80% disappearing within 30 min (Fig. [102]2E, F). W742L and W742C mutations showed no difference in their response to Bica/Enza-induced condensate formation/dissolution. Finally in this study, we engineer a streamlined phase separation tool ARDrop (AR-del (44-163, 226-347, DBD-H)-W742L/C), derived from the natural AR protein, that can be bi-directionally regulated by FDA-approved drugs. Time-lapse fusion experiments and FRAP (Fluorescence recovery after photobleaching) were performed to further confirm the liquid-like properties of ARDrop (Supplementary Fig. [103]4B and Supplementary Movie [104]1, Supplementary Fig. [105]4C and Supplementary Movie [106]2). ARDrop exhibited similar liquidity to AR-delNLS and AR-delDBD-H, with a t1/2 of ~6–8 s—shorter than WT (12 s) (Supplementary Fig. [107]1). Moreover, ARDrop showed no colocalization with stress granules (Supplementary Fig. [108]2) or protein degradation markers (e.g., Hsp70, p62, ubiquitin) (Supplementary Fig. [109]3), suggesting ARDrop condensates are distinct and independent from canonical membraneless organelles or insoluble degradation sites. To better illustrate the orthogonality of the use of androgens and antiandrogens, we monitored the dose-response of ARDrop to relevant ligands and then a comparison with the concentrations that are naturally found in cells/body. Using stable cell lines expressing ARDrop, we performed treatments with: DMSO (negative control), DHT (natural androgen), Bica (antiandrogen drug) and ARV-110 (PROTAC compound) across 10 concentration gradients (2-1024 nM) (Supplementary Fig. [110]4D). ARV-110, an orally bioavailable proteolysis-targeting chimera (PROTAC) targeting the androgen receptor, was used as a control to degrade AR and eliminate phase-separated condensates. ARDrop tool exhibits significant concentration-dependent condensates formation starting at 128 nM DHT, that are approximately two orders of magnitude higher than physiological androgen levels (0.1–3 nM in vivo^[111]23,[112]24). For the antiandrogen bicalutamide, while clinical serum concentrations reach 10−30 μM during treatment^[113]25, ARDrop only respond above 64 nM Bica - representing a 500-fold safety margin. These results collectively demonstrate that the ARDrop synthetic system remains unresponsive under physiological DHT concentrations, and is sensitive for bicalutamide at doses far below thearpeutic levels. Orchestrating kinase activity and signaling pathway using ARDrop The rapid induction and reversible nature of ARDrop suggested that it could be useful for manipulation of protein function and cellular processes. Cell signaling regulation coordinates key processes like proliferation, differentiation, survival, and metabolism, ensuring cellular and tissue function. Phase separation enriches signaling molecules to activate downstream pathways, as seen with LAT clustering in T-cell receptor signaling^[114]26, DNA-induced cGAS condensation in innate immune signaling^[115]27 and Rlα phase separation in cAMP signaling^[116]28. To test the use of ARDrop to manipulate function, we examined its ability to regulate the activity of anaplastic lymphoma kinase (ALK) protein, involved in cell growth and survival via key pathway like RAS-MAPK, with aberrant activation implicated in various cancers, including non-small cell lung cancer, lymphoma, and neuroblastoma. We fused ALK with ARDrop system (referred to as ARDrop-ALK) and time-lapse imaging revealed that phase-separated condensates appeared gradually 15 min upon bicalutamide treatment, while completely dissolved within 45 min after enzalutamide addition in HEK293 cells (Fig. [117]3A). FRAP experiments and time-lapse fusion experiments further confirmed the liquid-like properties of these synthetic condensates (Fig. [118]3B, Supplementary Movie [119]3, Supplementary Fig. [120]5A and Movie [121]4). Similar regulation of ALK kinase activity by ARDrop condensates can also be observed in HeLa and A375 cells (Supplementary Fig. [122]5B, C). Fig. 3. Inducible control of kinase activity and signaling pathway by ARDrop system. [123]Fig. 3 [124]Open in a new tab A Time-lapse imaging of HEK293 cells stably expressing ARDrop-ALK (green), treated sequentially with Bica followed by Enza. Scale bar, 40 µm. Statistical analysis of the LLPS puncta number of ARDrop-ALK (Right), n = 30 fields. B Image of FRAP experiments with ARDrop-ALK condensates (green) in ARDrop-ALK stably expressed HEK293 cells. Quantification of FRAP data are shown on the Right. Scale bar, 5 µm. Data are presented as mean ± SD of n = 7 independent experiments. C Schematic illustration of the recruitment of downstream signaling molecule GRB2 by ARDrop-ALK under the regulation of Bica and Enza. D Representative time-lapse imaging of HEK293T cells co-transfected with ARDrop-ALK (green) and GRB2-miRFP670 (red), treated sequentially with Bica followed by Enza. Scale bar, 40 µm. Quantification of fluorescence intensity of ARDrop-ALK (green) and GRB2-miRFP670 (red) along the line indicated in merged image were shown (Bottom). E Statistical analysis of the LLPS puncta of ARDrop-ALK (green) and GRB2- miRFP670 (red) at different time points following sequential treatment with Bica and Enza. Data are represented as mean ± SEM, n = 43 fields. F Schematic illustration of the recruitment of downstream signaling molecule SHC1 by ARDrop-ALK under the regulation of Bica and Enza. G Representative time-lapse imaging of HEK293T cells co-transfected with ARDrop-ALK (green) and SHC1-mScarlet (yellow), treated sequentially with Bica followed by Enza. Scale bar, 40 µm. Quantification of fluorescence intensity of ARDrop-ALK (green) and SHC1-mScarlet (yellow) along the line indicated in merged image were shown (Bottom). H Statistical analysis of the LLPS puncta of ARDrop-ALK (green) and SHC1-mScarlet (yellow) at different time points following sequential treatment with Bica and Enza. Data are represented as mean ± SEM, n = 87 fields. Box plots display the median (center line), interquartile range (IQR; box boundaries between 25th and 75th percentiles), whiskers (1.5 × IQR from the box), and outliers (points beyond 3 × IQR). Source data are provided in [125]Source Data file. Fig. 6. Manipulating cell apoptosis using ARDrop. [126]Fig. 6 [127]Open in a new tab A Time-lapse imaging of HEK293 cells stably expressing ARDrop-Caspase9 (green), treated sequentially with Bica followed by Enza. Scale bar, 40 µm. Statistical analysis of the LLPS puncta number of ARDrop-Caspas9 (Right), n = 238 fields. B Schematic diagram of cell morphological changes during apoptosis. C Time-lapse imaging of HEK293T cells transfected with ARDrop-Caspase9 (green), before and after the addition of Bica. Scale bar: 40 µm. White arrows indicate the apoptotic bodies. D Time-lapse imaging of HEK293 cells stably expressing ARDrop-Caspase9 (green) treated with Bica. “Bright” represents images captured in the bright field. “Merge” refers to the combination of bright field and GFP channels. Scale bar: 20 µm. White arrows indicate the apoptotic bodies. E Percentages of Annexin V-PE positive cells from gated cells. Statistical analysis of Annexin V-PE positive cells (Bottom). Data are presented as the mean ± SD of n = 3 independent experiments. p-values were calculated by one-way ANOVA. Box plots display the median (center line), interquartile range (IQR; box boundaries between 25th and 75th percentiles), whiskers (1.5 × IQR from the box), and outliers (points beyond 3 × IQR). Source data are provided in [128]Source Data file. To further dissect the precise regulation of ALK function by ARDrop tool, we spatiotemporally monitored the localization of signaling molecules that interact with ALK enabling downstream activation^[129]29,[130]30. Time-lapse imaging revealed that prior to bicalutamide treatment, ARDrop-ALK was diffusely distributed in the cytoplasm, while GRB2-miRFP670 was primarily nuclear but also present in the cytoplasm (Fig. [131]3C, D). Six minutes after bicalutamide addition, ARDrop-ALK began forming green condensates in the cytoplasm, which grew into larger droplets (Supplementary Movie [132]5). By ~7.5 min, GRB2-miRFP670 began shifting from the nucleus to the cytoplasm, gradually forming red condensates that colocalized with the green condensates (Fig. [133]3D, E). Notably, both the green and red droplets gradually increased in a time-dependent manner (Fig. [134]3E). After 22.5 min of bicalutamide treatment, enzalutamide was added, leading to gradual dissolution of both ARDrop-ALK and GRB2-miRFP670 droplets within 45 min (Fig. [135]3D, E, Supplementary Movie [136]5). Dynamic recruitment and release of signaling proteins SHC1 (tagged with mScarlet) as ARDrop-ALK condensates form and dissipate were also observed (Fig. [137]3F–H, Supplementary Movie [138]6), while cells transfected with GRB2-miRFP670 or SHC1-mScarlet alone exhibited no response to bicalutamide (Supplementary Fig. [139]5D, E). To validate ALK condensate specificity, we generated a kinase-dead mutant (ARDrop-ALK (K589M))^[140]29 and found that it abolished downstream recruitment of SHC1 and GRB2 within bicalutamide-induced condensates (Supplementary Fig. [141]6A, B, Supplementary Movies [142]7 and [143]8). Consistently, ALK kinase inhibitor Crizotinib also blocked SHC1/GRB2 recruitment (Supplementary Fig. [144]6C, D, Supplementary Movies [145]9 and [146]10). Notably, we observed that phospho-ERK, a marker of MAPK signaling activation, exhibited a first increasing and then decreasing pattern, in response to the bi-directional regulation of phase separation (Fig. [147]4A–E). A similar condensate-regulated kinase function was observed in the ARDrop chimera with another kinase, Neurotrophic Tyrosine Receptor Kinase (NTRK)^[148]31. The level of MAPK activation and signaling molecules recruitment correlate well with condensates formation (Supplementary Fig. [149]7 and Fig. [150]4F–J, Supplementary Movies [151]11 and [152]12). RNA-seq data indicated that while ARDrop-ALK overexpression increased ALK expression, it failed to activate downstream signaling without ligand (Fig. [153]4K, L, Supplementary Data [154]1) and corroborates that bicalutamide caused the hyperactivation of MAPK pathway in ARDrop-ALK system (Fig. [155]4M, N, Supplementary Data [156]2). We also performed qPCR experiments to validate that Enza effectively reverses bicalutamide-activated gene expression (Supplementary Fig. [157]8). These results demonstrate that ARDrop is effective in bidirectional regulating kinase activity and its downstream signaling. Fig. 4. RNA-seq analysis results of ARDrop-ALK. [158]Fig. 4 [159]Open in a new tab A Time-lapse imaging of HEK293 cells stably expressing ARDrop-ALK (green), treated sequentially with Bica followed by Enza. Scale bar, 40 µm, n = 142 fields. B LLPS puncta number quantification for (A). C Immunoblots of pErk and Erk in HEK293 cells stably expressing ARDrop-ALK, treated with Bica and Enza at indicated time point. D pErk/Erk band intensity ratios (normalized to blank) for (C). Data are presented as the mean ± SD of n = 3 independent experiments. p-values were calculated by one-way ANOVA. E Pearson’s correlation analysis (two-tailed) showing the correlation between band intensity ratios and the number of ARDrop-ALK puncta per cell. F Time-lapse imaging of HEK293 cells stably expressing ARDrop-NTRK3 (green), treated sequentially with Bica followed by Enza. Scale bar, 40 µm, n = 142 fields. G LLPS puncta number quantification for (F). H Immunoblots of pErk and Erk in HEK293 cells stably expressing ARDrop-NTRK3, treated with Bica and Enza at indicated time point. I pErk/Erk band intensity ratios (normalized to blank) for (H). Data are presented as the mean ± SD of n = 3 independent experiments. p-values were calculated by one-way ANOVA. J Pearson’s correlation analysis (two-tailed) showing the correlation between band intensity ratios and the number of ARDrop-NTRK3 puncta per cell. K Volcano plot of ARDrop-ALK vs blank HEK293 cells (|FC| > 2, p < 0.05; red/blue points: upregulated/downregulated mRNAs). Differential expression analysis is carried out using the DESeq2. L KEGG pathway enrichment analysis of differentially expressed genes between blank HEK293 cells and HEK293 cells stably expressing ARDrop-ALK, with significant categories (p < 0.05) exhibited. M Volcano plot of Bica- vs DMSO-treated HEK293 cells stably expressing ARDrop-ALK (|FC| > 2, p < 0.05; red/blue: up/down-regulated mRNAs). Differential expression analysis is carried out using the DESeq2. N KEGG pathway enrichment analysis of differentially expressed genes between Bica-treated and DMSO-treated ARDrop-ALK stably expressing HEK293 cells, highlighting the top 10 significant categories (p < 0.05). Box plots display the median (center line), interquartile range (IQR; box boundaries between 25th and 75th percentiles), whiskers (1.5 × IQR from the box), and outliers (points beyond 3 × IQR). Source data are provided in [160]Source Data file. Modulating stress granules dynamics using ARDrop Stress granules (SG) are membrane-less organelles that form in response to cellular stress, sequestering untranslated mRNAs and translation factors to conserve energy and prioritize stress-response protein synthesis^[161]32–[162]34. They maintain cellular homeostasis, and their dysregulation is linked to neurodegenerative disorders and cancers^[163]35,[164]36. Synthetic biology tools offering precise control over stress granule formation and disassembly will help researchers to study how aberrant granule dynamics contribute to diseases like ALS and Alzheimer’s and discover potential therapeutics. Prior to bicalutamide treatment, ARDrop-G3BP1 was diffusely distributed throughout the cytoplasm. After adding bicalutamide, green fluorescent droplets quickly appeared in the cytoplasm, subsequently enlarging into prominent green droplets within 15 min. These droplets were rapidly dissolved by enzalutamide within 45 min (Fig. [165]5A, Supplementary Movie [166]13). Next, we investigated the molecular players and downstream events of bica-enza regulated stress granule. Recent studies^[167]34,[168]37 have shown that G3BP1 plays a central role in stress granule (SG) assembly, acting as a scaffold protein for SG formation. Additionally, G3BP2 (Ras-GTPase-activating protein-binding protein 2) and PABPC1 (Poly(A)-binding protein cytoplasmic 1) are two key proteins associated with the formation and regulation of stress granules^[169]38,[170]39. As shown in Fig. [171]5B, C, about 9 min after treatment with bicalutamide, green G3BP1 droplets began to form, followed by the appearance of PABPC1 droplets that colocalized with G3BP1, behaving similarly to the endogenous G3BP1 condensates (Supplementary Fig. [172]9A)^[173]40. With longer bicalutamide treatment, both G3BP1 and PABPC1 droplets grew larger and more prominent. Quantitative analysis showed that G3BP1 and PABPC1 condensation reached half-maximal levels at ~15 and 20 min, respectively (Fig. [174]5D). After adding enzalutamide, time-lapse fluorescent imaging showed that G3BP1 and PABPC1 droplets gradually decreased, disappearing completely around 45 min after enzalutamide treatment (Fig. [175]5D, Supplementary Movie [176]14). Similar results were also observed with another crucial SG protein, G3BP2 (Fig. [177]5E–G, and Supplementary Movie [178]15). Cells only transfected with PABPC1-mScarlet or G3BP2-mScarlet exhibited no response to bicalutamide (Supplementary Fig. [179]9B, C). Bi-directional regulation of stress granules by ARDrop condensates can also be found in HeLa and A375 cells (Supplementary Fig. [180]9D, E). Researches have shown that persistent assembly of stress granules may lead to impaired cellular function and a decrease in cell viability^[181]41. Colony formation experiments using bicalutamide-treated ARDrop-G3BP1 cells revealed a time-dependent cell death induced by stress granule formation. 36 h after bicalutamide treatment, the number of colonies decreased by 50% (Fig. [182]5H), while bicalutamide itself did not affect the process (Supplementary Fig. [183]9F, G). These results indicate that ARDrop can modulate stress granule dynamics efficiently and will be useful to further characterizations for a mechanistic understanding of SG formation and composition. Fig. 5. Modulating stress granules dynamics using ARDrop. [184]Fig. 5 [185]Open in a new tab A Time-lapse imaging of HEK293 cells stably expressing ARDrop-G3BP1 (green), treated sequentially with Bica followed by Enza. Scale bar, 40 µm. Statistical analysis of the LLPS puncta number of ARDrop-G3BP1 (Right). Data are represented as mean ± SEM, n = 153 fields. B Schematic illustration of the recruitment of client protein PABPC1 by ARDrop-G3BP1 under the regulation of Bica and Enza. C Representative time-lapse imaging of HEK293T cells co-transfected with ARDrop-G3BP1 (green) and PABPC1-mScarlet (yellow), treated sequentially with Bica followed by Enza. Scale bar, 40 µm. Quantification of fluorescence intensity of ARDrop-G3BP1 (green) and PABPC1-mScarlet (yellow) along the line indicated in merged image were shown (Bottom). n = 104 fields. D Statistical analysis of the LLPS puncta of ARDrop-G3BP1 (green) and PABPC1-mScarlet (yellow) at different time points following sequential treatment with Bica and Enza. Data are represented as mean ± SEM, n = 104 fields. E Schematic illustration of the recruitment of client protein G3BP2 by ARDrop-G3BP1 under the regulation of Bica and Enza. F Representative time-lapse imaging of HEK293T cells co-transfected with ARDrop-G3BP1 (green) and G3BP2-mScarlet (yellow), treated sequentially with Bica followed by Enza. Scale bar, 40 µm. Quantification of fluorescence intensity of ARDrop-G3BP1 (green) and G3BP2-mScarlet (yellow) along the line indicated in merged image were shown (Bottom). n = 107 fields. G Statistical analysis of the LLPS puncta of ARDrop-G3BP1 (green) and G3BP2-mScarlet (yellow) at different time points following sequential treatment with Bica and Enza. Data are represented as mean ± SEM, n = 107 fields. H HEK293 cells stably expressing ARDrop-G3BP1 were treated with Bica for indicated times, and cell viability was assessed using crystal violet staining. Statistical analysis of the colony number (Right). Data are presented as the mean ± SD of n = 3 independent experiments. p-values were calculated by one-way ANOVA. Source data are provided in [186]Source Data file. Manipulating cell apoptosis using ARDrop The dimerization of caspase9 is a central step in initiating and driving the apoptosis process^[187]42. Following intrinsic apoptotic signals such as DNA damage or oxidative stress, caspase9 undergoes dimerization through interactions with various intracellular signaling molecules. This dimerization induces a conformational change in caspase9, exposing its active site, enabling it to cleave and activate downstream effector caspases (such as caspase3 and caspase7)^[188]43. These effector caspases then act on multiple cellular structures and functional proteins, leading to hallmark apoptotic features^[189]44. We sought to determine whether ARDrop can directly regulate cell apoptosis by fusing it with caspase9, independent of intrinsic apoptotic signals. Time-lapse results indicated that ARDrop was able to induce the condensate formation of caspase9 in cytoplasm within 15 min, and condensates induced by bicalutamide could be rapidly eliminated by enzalutamide within 30 min (Fig. [190]6A). During apoptosis, cells undergo characteristic morphological changes, including condensation, fragmentation, and the formation of apoptotic bodies (Fig. [191]6B). Time-lapse fluorescence imaging of HEK293T revealed that within 10 min of bicalutamide treatment, ARDrop-caspase9 condensates began to form, gradually increasing in size and brightness. About 30 min of bicalutamide exposure, cell condensation and fragmentation were observed, followed shortly by the appearance of apoptotic bodies (Fig. [192]6C, Supplementary Movie [193]16). Upon closer examination of the bicalutamide-induced ARDrop-caspase9 cell death process, HEK293 cells began to condense at 50 min, showed signs of fragmentation by 75 min, and eventually formed apoptotic bodies by 100 min (Fig. [194]6D, Supplementary Movie [195]17). We further analyzed bicalutamide-induced apoptotic cell death modulated by phase separation using Annexin V-PE staining and FACS. Approximately 40% of cells underwent apoptosis after 1 h of bicalutamide treatment, with nearly all cells becoming apoptotic after 3 h (Fig. [196]6E). Cell apoptosis regulated by ARDrop-caspase9 condensates can also be found in HeLa and A375 cells (Supplementary Fig. [197]10A–D). Manipulating cell necroptosis using ARDrop We also fused MLKL (mixed lineage kinase domain-like protein), a key executor of necroptotic cell death, with the ARDrop system to study its function. During necroptosis, MLKL oligomerizes and translocates to the plasma membrane, ultimately causing membrane rupture^[198]45,[199]46. Following bicalutamide treatment, MLKL-ARDrop puncta appeared on the cell membrane within 3 min, began to outline the membrane by 5 min, and showed clear accumulation along the membrane as cell death became evident after 10 min. The statistical results showed that about 10 min of bicalutamide treatment, the number of MLKL puncta peaked, followed by the onset of cell death. Both the green fluorescence intensity and the number of spots started to decline (Fig. [200]7A, Supplementary Movie [201]18). Notably, MLKL condensates cannot be reversed by enzalutamide. After just 2 min of bicalutamide treatment, MLKL condensates form on the cell membrane. Even when enzalutamide is subsequently added, the MLKL condensates persist, ultimately directing the cell toward necroptosis (Supplementary Fig. [202]11A, Supplementary Movie [203]19). To further confirm whether necroptosis occurs in the cells, we used PI (Propidium iodide) staining for detection and captured bright-field images to characterize the morphological changes. Bright-field observations revealed that at 5 min of bicalutamide treatment, the cells started to swell; by 12 min, bubbling appeared on the cell membranes, followed by organelle breakdown and release of intracellular contents (Supplementary Movie [204]20, Fig. [205]7B, C). Time-lapse fluorescence imaging results indicated that at 20 min of bicalutamide treatment, the PI fluorescence intensity began to show a marked increase, gradually rising as the treatment duration extended (Fig. [206]7D). Colony formation experiment showed that after just 10 min of bicalutamide treatment, approximately 50% of the cells had died (Fig. [207]7E). Necroptotic cell death regulated by MLKL-ARDrop can also be observed in HeLa and A375 cells (Supplementary Fig. [208]11B, C). These results indicates that ARDrop chemogenetic tool can be used to temporally control and dissect cell death process. Fig. 7. Manipulating cell necroptotic death using ARDrop. [209]Fig. 7 [210]Open in a new tab A Time-lapse imaging of HEK293 cells stably expressing MLKL-ARDrop (green) treated with Bica. Scale bar, 40 µm. Statistical analysis of the LLPS puncta number of MLKL-ARDrop (Right). Data are represented as mean ± SEM, n = 36 fields. B Schematic diagram of cell morphological changes during Necroptosis. C Time-lapse imaging of HEK293 cells stably expressing MLKL-ARDrop treated with Bica. Bright-field channel reveals necroptosis-associated morphological changes including membrane bubbling and cell swelling (indicated by white arrows), propidium iodide (PI) (red) channel identifies loss of membrane integrity as a cell death marker, mEGFP channel (green) tracks MLKL-ARDrop condensate formation and localization. Scale bar: 40 µm. White arrows indicate the bubbling. D Statistical results of mEGFP and PI intensity in HEK293 cells stably expressing MLKL-ARDrop before and after Bica treatment. Data are presented as the mean ± SEM, n = 36 fields. E HEK293 cells stably expressing MLKL-ARDrop were treated with Bica for indicated times, and cell viability was assessed using crystal violet staining. Statistical analysis of the colony number (Right). Data are presented as the mean ± SD of n = 3 independent experiments. p-values were calculated by one-way ANOVA. Source data are provided in [211]Source Data file. Discussion Building on our previous work, we characterize an orthogonal and minimal ARDrop platform of which phase separation can be bidirectional-controlled by the concentration and exposure time of clinically used bicalutamide and enzalutamide. Our results show that the ARDrop system, when combined with Bica/Enza, functions as a reversible molecular switch, offering precise temporal control of biological processes (Fig. [212]8). Fig. 8. Engineering strategy for functional AR synthetic condensates. [213]Fig. 8 [214]Open in a new tab Schematic represents the concept and development of ARDrop platform: We discovered the phase separation module form natural androgen receptor and paired with ligands combo to regulate condensates formation/dissolution. Effector proteins were further fused for bidirectional regulation of cellular functions. The methodologies described here complement and expand the growing toolkit of synthetic condensate systems. Most existing tools for regulating phase separation are based on well-established light-responsive proteins (e.g., Cry2)^[215]6–[216]8 or chemically induced oligomerization systems (e.g., FRB-rapamycin-RKBP)^[217]11–[218]13. A distinctive feature of the ARDrop system is its use of a phase-separation module derived from natural intrinsically disordered proteins (IDPs). Given the highly flexible nature of IDPs, techniques for studying their interactions are limited, and efforts to identify phase-separation modules from natural IDPs have been rare. Advantages of ARDrop system include the requirement of only a single scaffold protein, rapid kinetics of condensate formation and dissolution with FDA-approved drugs, and the easy, robust visualization of dynamic cellular control across multiple contexts in a qualitative manner. Notably, ARDrop system maintains stable liquid-like properties without transitioning to gel/solid states, enabling sustained cellular function modulation. For instance, bicalutamide induces a robust 6-8-fold phospho-Erk upregulation in ARDrop-kinase - completely reversible by enzalutamide (Fig. [219]4D, I)- and triggers near-complete (100%) necroptotic death within 40 min (Fig. [220]7C, D). Cross-species building blocks (e.g., proteins derived from plants or fungi) typically exhibit excellent orthogonality in mammalian cells, while mammalian proteins (such as BCL6 and the AR protein used in our study) often face inherent orthogonality limitations in mammalian systems. It’s worth mentioning that we have modified and optimized the AR protein to decouple its original function and phase separation ability, so as to reduce interference with AR function and relevant signaling pathways (Fig. [221]4K–N). ARDrop’s LLPS activation thresholds (≥128 nM DHT; >64 nM bicalutamide) are 100× higher (vs. physiological 0.1-3 nM DHT^[222]23,[223]24) and 500× lower (vs. therapeutic 10-30 μM bicalutamide^[224]25) respectively, suggesting biological inertness under physiological DHT and sensitiveness for bicalutamide at concentration far below therapeutic levels. Considering that condensation may be happening before large puncta are visible, we confirmed the orthogonality of ligands in sensitive tool like MLKL-ARDrop: DHT triggers MLKL membrane translocation and necroptosis at a threshold of 200 nM (~200× above physiological concentrations) and bicalutamide effectively induces this response at concentrations as low as 25 nM (~1000× below clinical therapeutic levels) (Supplementary Fig. [225]12). However, the use of antiandrogen as stimuli poses potential limitations on ARDrop platform, since low concentrations of antiandrogen may still have physiological effects on cells or in the body. Further work will be needed to assess the orthogonal drug-protein system. To demonstrate its utility, we tested this system in three biological contexts: receptor tyrosine kinase signaling, RNA-binding proteins and cell death regulatory proteins. The site of attachment of proteins to the ARDrop platform primarily depends on whether the fusion affects the protein’s functional activity. For example, MLKL’s membrane-pore formation depends on its N-terminal four-helix bundle domain (4HBD) (Supplementary Fig. [226]13A)^[227]47. In the ARDrop-MLKL fusion, bicalutamide induced membrane translocation and puncta formation, but ARDrop’s spatial blockade of the 4HBD disrupted pore formation, preventing necroptosis (Supplementary Fig. [228]13B). Conversely, the C-terminal fusion (MLKL-ARDrop) retained full functionality—mediating both puncta assembly necroptosis (Supplementary Fig. [229]13C). For kinase ALK and NTRK fusion sites, we adopted the natural attachment positions from oncogenic fusion proteins EML4-ALK^[230]48 and ETV6-NTRK3^[231]31. Compared with other antiandrogen combinations such as Bica/Apal (Apalutamide) and Bica/Daro (Darolutamide)^[232]22, Bica/Enza remains the most efficient regulatory pair for AR synthetic condensates (Supplementary Fig. [233]14). Altogether, ARDrop represents a robust tool for bi-directional control of condensates formation and can be exploited to regulate cell function. Implemented with disease-relevant proteins, the ARDrop system could allow investigation into how abnormal protein function contributes to disease and how therapeutic disruption could reverse pathological symptoms. Methods Cell line and Cell Culture All cell lines were obtained from the Cell Bank of the Shanghai Institute of Biochemistry and Cell Biology (SIBCB) and were authenticated by short tandem repeat (STR) profiling and mycoplasma detection. Cells were maintained accordingly to the guidance from American Type Culture Collection. Human HEK293 (female, SCSP-5209), HEK293T (female, SCSP-502), HEK293FT (female, SCSP-5212), A375 (female, SCSP-533) and HeLa (female, SCSP-504) cells were maintained in high-glucose Dulbecco’s Modified Eagle Medium (DMEM), supplemented with 10% (v/v) Fetal Bovine Serum (FBS), and antibiotics consisting of penicillin (100 units/mL) and streptomycin (100 µg/mL). All cells were cultured at 37 °C in a humidified atmosphere with 5% CO[2] and 95% air. Stable cell line construction To generate stable cells, retroviral infections were used. Briefly, HEK293FT cells were co-transfected with viral plasmids (list in Supplementary Data [234]3) and packing plasmids (gag-pol and VSVG) using jetPRIME (Polyplus, 117-15). The supernatant of cell culture medium was collected 48 h post-transfection, filtered through 0.45 µm filter (Millipore), and used to infect HEK293 cells. 48 h after infection, 2 ug/ml puromycin was added to culture media for selection and pool of cells were used for further experiments. Plasmids transfection HEK293T cells were transiently co-transfected with plasmids using jetPRIME (Polyplus, 117-15). The procedure was as follows: HEK293T cells were seeded at 70% confluency in 24-well glass-bottom plates (Cellvis, P24-1.5H-N) and cultured overnight. The next day, 50 µL of jetPRIME buffer was added to a 1.5 mL Eppendorf tube, followed by 0.5 µg of each plasmid (for a total of 1 µg or 1.5 µg). After mixing, 2 µL of jetPRIME transfection reagent was added, briefly vortexed for 1 s, and centrifuged. The mixture was incubated for 10 min and then added to the wells. 24 h after transfection, representative images of cells in 24-well glass bottom plate were captured using live cell imaging. Compounds Dihydrotestosterone was purchased from ApexBio (ApexBio, USA). Apalutamide, Bicalutamide, Darolutamide and Enzalutamide were purchased from Selleck (Selleck, USA), Crizotinib was purchased from MedChemExpress (MCE, USA). Compounds were dissolved in DMSO. All compounds used in this study are listed in Supplementary Data [235]3. High content image Cells were cultured on a 24-well glass bottom plate (Cellvis, P24-1.5H-N), treated sequentially with 200 nM bicalutamide followed by 2 µM enzalutamide. Images were captured using the Opera Phenix Plus High Content Screening system in confocal mode (Perkin Elmer) with a 63× water immersion objective (numerical aperture (NA) = 1.15), at 37 °C in a 5% CO[2] atmosphere. Immunofluorescence cell staining Cells were plated in a 96-well glass-bottom black plate (P96-1.5H-N, Cellvis) and fixed with 4% paraformaldehyde (PFA) for 15 min at room temperature. Following fixation, samples were washed three times with PBST (phosphate-buffered saline containing 0.05% Tween-20) and blocked for 30 min using QuickBlock Blocking Buffer for Immunol Staining (P0260, Beyotime). After another PBST wash, cells were incubated overnight at 4 °C with primary antibodies diluted in QuickBlock Primary Antibody Dilution Buffer (P0262, Beyotime). The following antibodies were used: Pol II S5P (Sigma, 04-1572, 1:400), H3K27ac (Abcam, ab4729, 1:400), G3BP1(Proteintech, 66486-1-lg, 1:400), PABPC1 (Proteintech, 10970-1-AP, 1:400), HSP70 (CST, #4873, 1:400), p62 (MBL, PM045, 1:400), Ubiquitin (CST, #3936, 1:400), Alexa Fluor 555 goat anti-mouse IgG (Thermo Fisher, A32727, 1:1000), Alexa Fluor 555 goat anti-rabbit IgG (Thermo Fisher, A32732, 1:1000), Alexa Fluor 555 goat anti-rat IgG (Thermo Fisher, A48263, 1:1000). Antibodies were obtained from commercial suppliers (see Supplementary Data [236]3 for details), including supplier, catalog number, clone name, and lot number. Fluorescence recovery after photobleaching (FRAP) The FRAP assay was performed utilizing the FRAP module of the Leica Stellaris 5 confocal microscope. A 488 nm laser beam was employed to bleach AR synthetic condensates, focusing on a defined circular region of interest with 100% laser power. Time-lapse images were captured at defined intervals. Fluorescence intensity measurements were conducted using LAS X (v3.7.4), and results are presented relative to pre-bleaching time points. The data were analyzed and graphed using GraphPad Prism (v10. 1.2). Image data analysis Harmony software (Perkin Elmer) was used to quantify puncta, combining built-in modules for cell segmentation, puncta identification, and measurement. First, we applied basic illumination correction to the images, defining the cell area based on absolute intensity thresholds. Individual cells were identified using the Find Cell module, with parameters adjusted according to each image type. Filters were then applied to remove cells with abnormal shapes or excessively high expression levels. The Find Spot module was used to detect all potential puncta, including low-contrast ones. Using the PhenoLOGIC module, we trained a classifier by manually labeling true and false puncta, ensuring a diverse range of spot intensities, sizes, and shapes for accurate classification. The properties of cells and puncta were exported, and statistical analysis and visualizations were performed in Python (v 3.13.0) and GraphPad Prism (v 10.1.2). RNA Sequencing For RNA sequencing, total RNA was first extracted from uninfected HEK293 cells (blank group), HEK293 cells stably expressing ARDrop-ALK, and HEK293 cells stably expressing ARDrop-ALK treated with 200 nM bicalutamide or DMSO for 2 h. Followed by an assessment of RNA quality using Agilent 2100 Bioanalyzer. Then the libraries were constructed using VAHTS Univeral V6 RNA-seq Library Prep kit according to the manufacturer’s instructions. The libraries were sequenced on an Illumina Novaseq 6000 platform and 150-bp paired-end reads were generated. During data preprocessing with Trimmomatic, clean reads are obtained by removing reads with adapters, poly-N, or low quality. These reads are then aligned to the human genome (GRCh38) using HISAT2. For mRNA analysis, gene expression levels are calculated as FPKM using Cufflinks, and read counts are generated with HTSeq-count. Differential expression analysis is carried out using the DESeq2 (v1.42.1), setting a significance threshold at P-value < 0.05 and fold change >2. The volcano plot was generated using the EnhancedVolcano R package (v1.20.0) and GO and KEGG analyses were conducted using the clusterProfiler R package (v4.10.1). RNA extraction and qRT-PCR Total RNA was isolated from whole-cell lysates using TRIzol reagent (15596018CN; Invitrogen), with RNA integrity verified by 1% agarose gel electrophoresis. cDNA was synthesized from 1 μg total RNA using the PrimeScript™ RT Reagent Kit (RR0037A; Takara). Quantitative PCR was performed in triplicate using TB Green® Premix Ex Taq™ (RR420Q; Takara) on a QuantStudio 6 Real-Time PCR System (Applied Biosystems). Relative mRNA expression levels were determined via the 2^−ΔΔCT method using GAPDH as the endogenous control. Primer sequences are detailed in Supplementary Data [237]4. Western blot Uninfected HEK293 cells (blank group), HEK293 cells stably expressing ARDrop-ALK or ARDrop-NTRK3 (either untreated or treated with 200 nM bicalutamide / 2 µM enzalutamide or DMSO for the specified duration), were collected and lysed in ice-cold RIPA lysis buffer (P0013B, Beyotime), which was supplemented with phosphatase and protease inhibitors, and incubated on ice for 10 min. The resulting supernatant was used for protein concentration determination via the BCA protein quantification kit (P0009, Beyotime). A total of 20 µg of the extracted protein was loaded onto an SDS-PAGE gel for separation and subsequently transferred to nitrocellulose membranes (Whatman). The membranes were blocked with 5% bovine serum albumin (BSA) in TBST before being incubated with primary antibodies, followed by the appropriate secondary antibodies. Finally, the specific protein bands on the membranes were visualized using Typhoon NIR imager from Cytiva. The following antibodies were used: p44/42 MAPK (Erk1/2) (CST#9102, 1:1000), Phospho-p44/42 MAPK (Erk1/2) (Thr202/Tyr204) (CST#4370, 1:1000), IRDye 800CW Goat anti-Rabbit IgG (H + L) (LI-COR Biosciences #926-32211, 1:1000), IRDye 680RD Goat anti-Mouse IgG (H + L) (LI-COR Biosciences#926-68170, 1:1000). Antibodies were obtained from commercial suppliers (see Supplementary Data [238]3 for details), including supplier, catalog number, clone name, and lot number. Colony formation assay Both ARDrop-Caspase9/MLKL-ARDrop stably expressing cells and blank HEK293 control cells were seeded in 12-well plates at a density of 1000 cells per well. The following day, the cells were treated with bicalutamide for specified duration. After treatment, the medium containing bicalutamide was replaced with fresh complete medium, and the cells were cultured for an additional 10−14 days. Subsequently, the colonies were washed with PBS, fixed with 95% ethanol, and stained with 0.1% crystal violet. Finally, the colonies were imaged and counted. Flow cytometry The apoptosis analysis was performed using Annexin-V-PE (C1065, Beyotime) according to the manufacturer’s protocol. Briefly, cells were collected and washed with cold PBS twice, and then were stained with Annexin V-PE in binding buffer at room temperature for 15 min in the dark. Finally, the percentage of apoptotic cells was quantified using the Cytek Aurora flow cytometer (Cytek Biosciences), with data analysis performed in FlowJo software (v10.8.1). The gating strategy proceeded as follows (Supplementary Fig. [239]15): First, intact cells were selected by FSC-H/SSC-H characteristics, followed by doublet exclusion using FSC-A/FSC-H parameters. Apoptotic cells (Annexin V-PE+) were then resolved from viable cells (Annexin V-PE-) within this singlet population. Statistics and reproducibility All statistical analyses were performed using GraphPad Prism (v10.1.2; GraphPad Software, La Jolla, CA) and Python (v3.13.0). Data are presented as mean ± standard error of the mean (SEM) or mean ± standard deviation (SD), as indicated, with a minimum of three independent experimental replicates. Sample sizes for each experiment are detailed in the corresponding figure legends. For high-content imaging analyses, the number of fields captured per sample is specified in the respective figure legends. Box plots were generated using Python (v3.13.0) with pandas (v2.3.0) for data processing and matplotlib (v3.10.3) for visualization. Box plots display the median (center line), interquartile range (IQR; box boundaries between 25th and 75th percentiles), whiskers (1.5 × IQR from the box), and outliers (points beyond 3 × IQR). For group comparisons, unpaired two-tailed Student’s t-tests were used for two groups, while one-way ANOVA with Dunnett’s post hoc test was applied for three or more groups, with significance thresholds set at *p < 0.05, **p < 0.01, ***p < 0.001, and ****p < 0.0001. No statistical method was used to predetermine sample size. No data were excluded from the analyses. All samples were randomly grouped. The Investigators were not blinded to allocation during experiments and outcome assessment. Reporting summary Further information on research design is available in the [240]Nature Portfolio Reporting Summary linked to this article. Supplementary information [241]Supplementary Information^ (5MB, pdf) [242]41467_2025_61877_MOESM2_ESM.docx^ (19KB, docx) Description of Additional Supplementary Information [243]Supplementary Data^ (30.4KB, xlsx) [244]Supplementary Movie 1^ (2.8MB, avi) [245]Supplementary Movie 2^ (28.3MB, avi) [246]Supplementary Movie 3^ (21.6MB, avi) [247]Supplementary Movie 4^ (1.2MB, avi) [248]Supplementary Movie 5^ (2.9MB, avi) [249]Supplementary Movie 6^ (2.7MB, avi) [250]Supplementary Movie 7^ (21.2MB, avi) [251]Supplementary Movie 8^ (28.1MB, avi) [252]Supplementary Movie 9^ (16.2MB, avi) [253]Supplementary Movie 10^ (17.8MB, avi) [254]Supplementary Movie 11^ (2.9MB, avi) [255]Supplementary Movie 12^ (3MB, avi) [256]Supplementary Movie 13^ (22.2MB, avi) [257]Supplementary Movie 14^ (3.7MB, avi) [258]Supplementary Movie 15^ (4.2MB, avi) [259]Supplementary Movie 16^ (10.8MB, avi) [260]Supplementary Movie 17^ (10.3MB, avi) [261]Supplementary Movie 18^ (1MB, avi) [262]Supplementary Movie 19^ (2.2MB, avi) [263]Supplementary Movie 20^ (15.4MB, avi) [264]Reporting Summary^ (92.1KB, pdf) [265]Transparent Peer Review file^ (8.9MB, pdf) Source data [266]Source Data^ (9.1MB, xlsx) Acknowledgements