Pharmacologic inhibition of PCBP2 biomolecular condensates relieves Alzheimer's disease

Abstract

Biomolecular condensates, membrane-less assemblies formed by phase separation, are implicated in neurodegenerative disease, but their role in Alzheimer's disease (AD) remains unclear. Here, we report that in the brain of AD patients and animal models, an elevation of poly(C)-binding protein 2 (PCBP2) correlates with biomolecular condensation that involves phase separation. These condensates sequester large numbers of mitochondrial and mRNA-binding proteins, leading to the outside impairment of mitochondrial morphology and function, and BACE1 mRNA decay relative to amyloid deposition. We then identify a small molecule CN-0928 that inhibits the condensates by reducing PCBP2 protein level and mitigates AD pathology and cognitive decline, in which CN-0928 binding to a target protein integrator complex subunit 1 (INTS1) allows to regulate PCBP2 expression. Our findings place PCBP2 condensates as a key player that cooperates the seemingly disparate but important pathways, and show pharmacological modulation of PCBP2 as an effective approach for treating AD.

Fig. 1. PCBP2 condensates correlate with increased PCBP2 protein abundance in AD.

a Western blots (left) and quantification (right) of PCBP2 and APP in the temporal cortex of postmortem brains from both non-AD subjects (Ctrl, C1-C10) and AD patients (AD, P1-P10). b Representative immunofluorescence images (left) and quantification (right) of PCBP2 condensates (red) in the temporal cortex from non-AD controls and AD patients. NeuN (green) and DAPI (blue) are counterstained. Scale bar, 10 μm or as indicated. c, d Western blots (left) and quantification (right) of PCBP2 and APP protein levels in the hippocampus (c) and cortex (d) of 5×FAD and wild-type (WT) mice. e Representative immunofluorescence images (left) and quantification (right) showing PCBP2 condensates (red) in the hippocampus of 12-month-old WT and 5×FAD mice. Scale bar, 10 μm. f Representative immunofluorescent images of PCBP2 (red) in SH-SY5Y, SH-SY5Y-APP, or SH-SY5Y cells treated with arsenite (Ars, 0.5 mM for 30 min). Scale bar, 10 μm. g Live-cell imaging of SH-SY5Y cells stably expressing mCherry-PCBP2 (SH-SY5Y-mCherry-PCBP2), counterstained with Hoechst (blue). Scale bar, 10 μm. h Time-lapse confocal fluorescence imaging of SH-SY5Y cells stably expressing mCherry or mCherry-PCBP2, following treatment with 1.5% 1,6-hexanediol (1,6-HD). Scale bar, 10 μm. Data are expressed as mean ± s.e.m (a–e). Significance was determined by the unpaired two-tailed Student’s t-test (a–e). In vitro: n = 3 biological replicates (f–h). In vivo: n = 6 mice/group (c–e); n = 7 samples (b [Ctrl group]); n = 8 samples (b [AD group]); n = 10 samples (a). Source data are provided as a Source Data file.

Fig. 2. PCBP2 forms liquid-like condensates via LLPS.

a Phase separation of the full-length PCBP2 (PCBP2-A555, 5 μM) occurs in 10% PEG, forming liquid droplets that are visualized by the differential interference contrast (DIC) channel and the fluorescent channel of confocal microscopy. The middle panel shows an isolated single small condensate, whereas the bottom panel shows a cluster of condensates. Scale bar, 1 μm. b Fluorescence recovery after photobleaching (FRAP) of PCBP2-A555 droplets was assessed in 10% PEG. The dashed box indicates the photobleached region (region of interest, ROI). Scale bar, 2.5 μm. c Representative confocal images of liquid-like droplet formation by PCBP2 in vitro at fixed PCBP2 (2 μM) with RNA at 0, 10, 20, and 50 ng/μL. Scale bar, 1 μm. d Representative confocal images at fixed RNA (20 ng/μL) with PCBP2 at 0, 1, 2, and 4 μM. Scale bar, 1 μm. e mCherry-PCBP2 condensates are observed in SH-SY5Y cells at 48 h post-transfection but not at 24 h. Representative fluorescence images (left) and FRAP recovery curve (right) at 48 h illustrate fluorescence recovery within condensates; the dashed box marks the photobleached region of interest (ROI). Scale bar, 10 μm. Data are presented as mean ± SD (b, e). In vitro: n = 3 biological replicates (a–e). Source data are provided as a Source Data file.

Fig. 3. PCBP2 condensates sequester mitochondrial and RNA-binding proteins.

a Isolation and analysis workflow for the characterization of PCBP2 condensates by fluorescence-activated particle sorting (FAPS). b–e Comparison of protein abundances between sorted and pre-sorted fractions utilizing quantitative values from normalized total spectra of LC-MS/MS. Significantly enriched and depleted proteins are indicated, alongside known p-body proteins (b), stress granule (SG) proteins (c), ribosomal proteins (d), and mitochondrial proteins (e). f Proteins identified by mass spectrometry were analyzed by gene ontology using the clusterProfiler with associated protein numbers (counts), fold enrichment (FE) values, and Q-values. g Representative images of mCherry-PCBP2 condensates colocalized with the indicated organelle-associated proteins by confocal sectioning. Scale bar, 10 μm. h No droplet is found when purified recombinant human TOM20-D488 (0.5 or 1 µM) is incubated with 10% PEG or 50 ng/µL RNA as indicated. Co-incubation of purified PCBP2-A555 (2 µM) with TOM20-D488 (0.5 µM) in buffers containing 20 ng/μL RNA yields a fluorescent droplet. Scale bar, 1 μm. In vitro: n = 3 biological replicates (g, h). Source data are provided as a Source Data file.

Fig. 4. PCBP2 condensates disrupt mitochondrial morphology and function.

a GSEA comparing pre- and post-sorted PCBP2-condensate fractions used a one-sided enrichment test; Multiple testing was controlled across gene sets using the Benjamini–Hochberg FDR. b Live-cell imaging (upper) of SH-SY5Y-PCBP2 cells stained with MitoTracker Green. Quantitative analysis (lower) of mitochondrial morphology is measured by area, perimeter, aspect ratio (AR), and form factor (FF) values using a computer-assisted morphometric system. Scale bar, 10 µm. c Immunofluorescence images (upper) of SH-SY5Y-PCBP2 cells stained for TOM20 (green), DAPI (blue), and PCBP2 (red). These descriptors (lower) also included area, perimeter, AR, and FF. Scale bar, 10 µm. d Transmission electron microscopy (TEM) images show an impairment of mitochondrial morphology in SH-SY5Y-PCBP2 cells. Mitochondria indicated by the white arrows in the left panels are displayed with higher magnification in the right panels. Black arrows indicate damaged mitochondria with ridge reduction or absence. Scale bar, 2 µm or as indicated. Mitochondrial (mito) mean number and length were quantified. e Live-cell imaging of reactive oxygen species (ROS) levels in SH-SY5Y cells transiently transfected with vector (Vector), PCBP2, or incubated with 50 µM hydrogen peroxide (H₂O₂) for 4 h without transfection. ROS was labeled with the DCFH-DA probe. Scale bar, 10 µm. f Oxygen consumption rate (OCR) between SH-SY5Y and SH-SY5Y-PCBP2 cells is assessed using a Seahorse XF24 Extracellular Flux Analyzer, respectively. Statistical results are presented as bar graphs at the bottom. g Extracellular acidification rate (ECAR) between SH-SY5Y and SH-SY5Y-PCBP2 cells is assessed using a Seahorse XF24 Extracellular Flux Analyzer, respectively. Statistical results are presented as bar graphs at the bottom. h Immunofluorescence images showing the colocalization of TOM20 (green) and PCBP2 (red) in SH-SY5Y-APP cells. Scale bar, 10 µm. PCC: Pearson’s correlation coefficient. i Live-cell imaging of reactive oxygen species (ROS) levels in SH-SY5Y-APP cells transiently transfected with siPCBP2 or a negative control (NC). ROS levels were visualized with the DCFH-DA probe. Scale bar, 10 µm. Data: mean ± SD (b–d, f–h) and analyzed by two-tailed Student’s t-test (b–d, f–h). In vitro: n = 3 biological replicates (b–e, f [SH-SY5Y], h, i); n = 4 biological replicates (f [SH-SY5Y-PCBP2]); n = 10 biological replicates (g). Source data are provided as a Source Data file.

Fig. 5. PCBP2 condensates inhibit the 3′UTR degradation of BACE1 mRNA.

a Scatter plot of differentially expressed genes in SH-SY5Y cells shows BACE1 transcript is downregulated by PCBP2 knockdown. b Western blots and quantification indicate that BACE1 protein is reduced by PCBP2 knockdown. c Actinomycin-D chase (0–10 h) shows faster decay of BACE1 mRNA with siPCBP2. d Dual-luciferase assays 48 h post-transfection with overexpression PCBP2 (oe-PCBP2) and reporters with or without 3′/5′UTRs reveal repression through the 3′UTR. e Schematic diagram of BrU-labeled 3′UTR RNA pulldown followed by LC-MS/MS to identify 3′UTR-binding RBPs. f Western blots of PCBP2 in immunoprecipitated extracts by RNA (BACE1 3′UTR) pulldown. g MS2-Trap and western blot confirm PCBP2 associates with the 3′UTR but not the 5′UTR; 3′UTRΔ, deletion control. h Map of the full-length BACE1 3′UTR truncated into four segments for reporter testing. i Relative luciferase activities of the truncated 3′UTRs reporter plasmids. j Schematic illustration shows nonsense-mediated mRNA decay (NMD) involves UPF1 and SMG1. k, l siPCBP2 lowers BACE1 mRNA (k) and protein (l); SMG1 inhibitor (SMG1i, 300 nM, 4 h) mitigates these effects; p-UPF1/UPF1 blots shown. m UPF1 knockdown rescues the siPCBP2-induced reduction in BACE1 mRNA. n Immunofluorescence of SH-SY5Y and SH-SY5Y-PCBP2 cells shows colocalization of PCBP2 (red), UPF1 (magenta), and BACE1 3′UTR (green, dEcCas6 probe); line scans along the yellow ROI; brown arrows indicate regions where the 3′UTR is localized with low UPF1 density. Scale bar: 10 μm. o PCBP2 domain map highlighting four predicted intrinsically disordered regions (IDR1-IDR4). p Confocal images of PCBP2-mCherry clusters formed by the indicated deletion mutants lacking IDR1-IDR4 (del IDR1-4) transiently expressed in SH-SY5Y cells. Scale bar, 2.5 µm. q Representative western blots and corresponding quantification of BACE1 and the PCBP2-mCherry deletion mutants (del IDR1-4) in SH-SY5Y cells. Data: mean ± s.e.m (b, k–m, q) and mean ± SD (c, d, i). Significance was determined by the unpaired two-tailed Student’s t-test (c, d, i), one-way ANOVA with Bonferroni’s test (b, k–m, q). ns: nonsignificant. In vitro: n = 3 biological replicates (b, c, f, g, k–n, p, q); n = 7 biological replicates (i [4^th-3′UTR oe-PCBP2 group]); n = 8 biological replicates (d, i [other groups]). Source data are provided as a Source Data file.

Fig. 6. Small-molecule CN-0928 inhibits condensates by reducing PCBP2 levels and alleviates AD.

a Chemical Structure of CN-0928. b–e Cells were treated with either DMSO or CN-0928 (1 μM) for 48 h. b Western blots (left) and quantification (right) of PCBP2, BACE1, APP, and ADAM10 in SH-SY5Y cells treated with CN-0928. c Live-cell imaging (left) and quantification (right) of PCBP2 condensates in SH-SY5Y-mCherry-PCBP2 cells incubated with CN-0928. Scale bar: 10 μm. d Western blots (left) and quantification (right) of α/β-CTFs. s.e. short exposure; l.e., long exposure. e Aβ40 and Aβ42 levels in cell lysates (in) and culture medium (ex) were measured by ELISA. f–p Male 5×FAD mice were injected with either normal saline (NS) or CN-0928. Wild-type (WT) mice received NS as the complete blank control. f, g Representative western blots (left) and quantification (right) of PCBP2, BACE1, and APP levels in the hippocampus (f) and cortex (g). h Representative immunofluorescence images (left) and quantification (right) of PCBP2 condensates in the hippocampus. Scale bars, 10 μm. i, j Representative western blots (left) and quantification (right) of α/β-CTFs levels in the hippocampus (i) and cortex (j). k Immunofluorescent labeling (upper) and quantification (lower) of Aβ deposits (green) in the hippocampus. Scale bar: 200 µm. l Soluble and insoluble Aβ40 and Aβ42 in the hippocampus were quantified using ELISA. m Representative trajectory maps for three groups of mice in the probe trials. n 5×FAD mice treated with CN-0928 exhibit a shorter escape latency compared with NS. o, p Passing times (o) and the staying time (p) in the target quadrant in different groups of mice. Data: mean ± s.e.m (b, f, g, i, j, l) and mean ± SD (c–e, h, k, n–p). Significance was determined by the unpaired two-tailed Student’s t-test (b–e, k, l), one-way ANOVA with Bonferroni’s test (g, i, j [α-CTF], n, p), Welch’s one-way ANOVA with Dunnett’s T3 post hoc test; Brown-Forsythe variance test (f, j [β-CTF], o), and Kruskal–Wallis test with Dunn’s post-test (h). ns: nonsignificant. In vitro: n = 3 biological replicates (c–e); n = 6 biological replicates (b). In vivo: n = 6 mice/group (h, l); n = 10 mice/group (f, g, i–k, n–p). Source data are provided as a Source Data file.

Fig. 7. CN-0928 downregulates PCBP2 protein by INTS1.

a The workflow of chemical proteomics for target identification of CN-0928 includes three groups: In the CN-0928 group (middle), SH-SY5Y cell lysate was incubated with CN-0928-conjugated beads to enrich putative CN-0928 target proteins. In the competition group (lower), the lysate was preincubated with excessive free CN-0928 and then incubated with CN-0928-conjugated beads. In the blank control (upper), beads without CN-0928 were used to identify nonspecifically bound proteins. Each experiment was performed in duplicate, and the proteins pulled down were analyzed by LC-MS/MS. b, c Representative Western blots (left) and quantification (right) of BACE1 and PCBP2 in SH-SY5Y cells with INTS1 (b) or mH2A1 (c) knockdown, followed by treatment with either 1 μM CN-0928 or DMSO (Ctrl). d Relative mRNA levels of PCBP2 and BACE1 in SH-SY5Y cells transiently transfected with siINTS1, measured in the absence (Ctrl) or presence of 1 µM CN-0928. e Molecular docking results of CN-0928 to INTS1. The panoramic view shows the spatial position of CN-0928 in the structure of INTS1. The hydrogen bond is presented in the local interaction image. f Per-residue decomposition analysis of the binding free energy. In the CN-0928-INTS1 binary complex, CN-0928 forms hydrogen bonds with ARG-1404 during molecular dynamics simulations. Violin width is proportional to the kernel-density estimate of the distribution; the central line denotes the median and the thin lines indicate the 25th and 75th percentiles (n = 1000 analyzed frames). g CN-0928-biotin pulldown of flag-tagged wild type (INTS1-WT-flag) and ARG-1404 mutated INTS1 (R1404A-flag, R1404K-flag, and R1404L-flag) proteins. Cell lysates were mixed without and drug (blank), or with CN-0928-biotin (CN-biotin). Data: mean ± s.e.m (b–d). Significance was determined by one-way ANOVA with Bonferroni’s test (b–d). ns nonsignificant. In vitro: n  =  3 biological replicates (b–d, g). Source data are provided as a Source Data file.