Huntington disease phase separation: mHTT low-complexity domain undergoes LLPS forming gel-like condensates that trap transcription factors — condensate-dissolving compounds could restore gene expression programs

The low-complexity domain (LCD) of mutant Huntingtin protein (mHTT), specifically the polyglutamine (polyQ) expansion region (>36 repeats) combined with the N17 amphipathic helix and proline-rich domain, undergoes liquid-liquid phase separation (LLPS) in neuronal cells to form gel-like condensates with material properties distinct from liquid droplets (higher viscosity η > 10³ Pa·s, reduced fluorescence recovery in FRAP < 20% within 60 minutes). These condensates sequester critical transcription factors (SP1, TFIID, CBP/p300, NF-Y) at concentrations ≥3-fold above cytoplasmic baseline, reducing their nuclear availability by ≥40% and suppressing downstream gene expression programs (BDNF, PGC-1α, dopamine receptor genes) by ≥50%. Small molecules capable of dissolving or fluidizing these condensates (reducing viscosity by ≥50%, restoring FRAP recovery to >60% within 30 minutes) will restore transcription factor availability and rescue gene expression to ≥70% of wild-type levels in HD model neurons.

PHASE 1 — In Vitro Biophysical Characterization (Weeks 1–6): Purify recombinant mHTT-N-terminal fragments (exon1, Q46, Q72, Q97) with GFP/mCherry tags. Perform turbidity assays (OD600 vs. concentration), DLS (hydrodynamic radius), and microrheology (passive bead tracking) to establish phase diagrams. Conduct FRAP on condensates formed at 2 μM protein, 37°C, pH 7.4 with 10% PEG-8000. Measure half-time of recovery and mobile fraction.

PHASE 2 — Transcription Factor Sequestration (Weeks 4–10): Co-incubate purified mHTT condensates with recombinant SP1-DBD, CBP-KIX, TFIID-TAF1 at equimolar ratios (1 μM each). Quantify partitioning by centrifugation (16,000×g, 10 min) and SDS-PAGE/western blot of pellet vs. supernatant fractions. Validate by fluorescence microscopy (confocal, TIRF).

PHASE 3 — Cellular Validation (Weeks 6–16): Use STHdh Q7/Q7 vs. Q111/Q111 striatal cell lines and iPSC-derived HD neurons (Q60, Q109). Perform: (a) live-cell imaging with mHTT-GFP and transcription factor-mCherry fusions; (b) FRAP in cells; (c) nuclear/cytoplasmic fractionation + western blot; (d) RNA-seq for transcriptome-wide expression changes.

PHASE 4 — Compound Screening and Rescue (Weeks 12–24): Screen 500–2,000 compounds (FDA-approved library + condensate-targeting compounds: 1,6-hexanediol analogs, small molecule chaperones, PPI inhibitors) using high-content imaging (condensate number, size, FRAP recovery as readouts). Top 20 hits validated by RNA-seq rescue assay.

PHASE 5 — In Vivo Validation (Weeks 20–36): Administer top 3 compounds to R6/2 mice (n=15/group) via IP injection or CNS delivery. Assess: striatal mHTT condensate properties by cryo-ET on brain slices, transcription factor nuclear availability by IF, gene expression by qRT-PCR/RNA-seq, behavioral outcomes (rotarod, open field).

# Experimental Validation Pipeline — HD Condensate EVP
# Pseudocode architecture for computational and data analysis components

class HDCondensateValidationPipeline:
    
    def __init__(self):
        self.protein_sequences = load_HTT_sequences(Q_lengths=[23, 46, 72, 97])
        self.transcription_factors = ['SP1', 'CBP', 'TFIID_TBP', 'NFY_A']
        self.compound_library = load_compound_library(n=2000)
        self.cell_lines = ['STHdh_Q7', 'STHdh_Q111', 'iPSC_Q60', 'iPSC_Q109']
    
    # MODULE 1: Phase Diagram Computation
    def compute_phase_diagram(self, sequence, conditions):
        """
        Coarse-grained MD using CALVADOS2 force field
        Input: HTT LCD sequence, salt/temperature conditions
        Output: Csat, phase boundary, condensate properties
        """
        model = CALVADOS2(sequence)
        trajectories = model.simulate(
            n_chains=100,
            box_size=50,  # nm
            temperature=310,  # K
            ionic_strength=0.15,  # M
            duration=10e-6,  # 10 microseconds
            timestep=10e-15  # 10 fs
        )
        Csat = calculate_critical_concentration(trajectories)
        contact_map = compute_contact_map(trajectories, cutoff=0.8)  # nm
        return PhaseData(Csat=Csat, contacts=contact_map, Rg=calc_Rg(trajectories))
    
    # MODULE 2: FRAP Analysis
    def analyze_FRAP(self, fluorescence_timeseries, bleach_time, pixel_size):
        """
        Fit FRAP recovery curves to extract mobile fraction and t_half
        """
        # Normalize: F_norm = (F(t) - F_bleach) / (F_pre - F_bleach)
        F_norm = normalize_FRAP(fluorescence_timeseries, bleach_time)
        
        # Single exponential fit: F(t) = A*(1 - exp(-t/tau)) + C
        params = curve_fit(
            lambda t, A, tau, C: A*(1-np.exp(-t/tau))+C,
            time_axis, F_norm,
            p0=[0.8, 30, 0.1],
            bounds=([0,0,0],[1,3600,0.5])
        )
        mobile_fraction = params['A']
        t_half = params['tau'] * np.log(2)
        
        # Classification
        if mobile_fraction > 0.8 and t_half < 30:
            state = 'LIQUID'
        elif mobile_fraction < 0.2 and t_half > 600:
            state = 'GEL'
        else:
            state = 'VISCOELASTIC'
        
        return FRAPResult(mobile_fraction=mobile_fraction, t_half=t_half, state=state)
    
    # MODULE 3: Transcription Factor Partitioning
    def calculate_partition_coefficient(self, pellet_intensity, supernatant_intensity,
                                         pellet_volume=0.01, total_volume=0.1):
        """
        Kp = ([TF]_condensate) / ([TF]_dilute_phase)
        """
        conc_condensate = pellet_intensity / pellet_volume
        conc_dilute = supernatant_intensity / (total_volume - pellet_volume)
        Kp = conc_condensate / conc_dilute
        return Kp  # Target: Kp >= 3.0
    
    # MODULE 4: RNA-seq Differential Expression
    def run_differential_expression(self, count_matrix, metadata):
        """
        DESeq2 wrapper for HD vs. WT comparison
        """
        dds = DESeqDataSet(count_matrix, design='~condition')
        dds = DESeq(dds)
        results = results(dds, contrast=['condition', 'Q111', 'Q7'],
                         alpha=0.05, lfcThreshold=1.0)
        
        # Filter significant genes
        sig_genes = results[results['padj'] < 0.05]
        suppressed = sig_genes[sig_genes['log2FoldChange'] < -1]
        
        # GSEA on ranked gene list
        ranked_genes = results.sort_values('stat', ascending=False)
        gsea_results = run_GSEA(ranked_genes, gene_sets=['HALLMARK', 'KEGG', 'HD_SIGNATURES'])
        
        return DEResults(suppressed_genes=suppressed, gsea=gsea_results,
                        n_DE=len(sig_genes))
    
    # MODULE 5: Compound Screening Analysis
    def analyze_compound_screen(self, plate_data, controls):
        """
        High-content imaging analysis for condensate dissolution
        """
        # Calculate Z-scores per plate
        pos_ctrl_mean = np.mean(controls['positive'])
        pos_ctrl_std = np.std(controls['positive'])
        
        for compound in plate_data:
            compound['condensate_area_zscore'] = (
                (compound['mean_condensate_area'] - pos_ctrl_mean) / pos_ctrl_std
            )
            compound['frap_recovery_zscore'] = (
                (compound['frap_mobile_fraction'] - pos_ctrl_mean) / pos_ctrl_std
            )
            # Combined score
            compound['hit_score'] = (
                compound['condensate_area_zscore'] + compound['frap_recovery_zscore']
            ) / 2
        
        # Select hits: Z-score < -2, viability > 70%
        hits = [c for c in plate_data 
                if c['hit_score'] < -2 and c['viability'] > 0.70]
        
        return sorted(hits, key=lambda x: x['hit_score'])
    
    # MODULE 6: In Vivo Analysis
    def analyze_mouse_behavior(self, rotarod_data, groups):
        """
        Statistical analysis of behavioral outcomes
        """
        # Rotarod: latency to fall (seconds)
        anova_result = one_way_ANOVA(rotarod_data, groups)
        posthoc = tukey_hsd(rotarod_data, groups)
        
        # Effect size (Cohen's d) for compound vs. vehicle in R6/2
        d = cohens_d(
            rotarod_data['R6/2_compound'],
            rotarod_data['R6/2_vehicle']
        )
        
        return BehaviorResult(anova=anova_result, posthoc=posthoc, effect_size=d)
    
    # MAIN EXECUTION PIPELINE
    def run_full_validation(self):
        # Phase 1: Biophysics
        phase_diagrams = {Q: self.compute_phase_diagram(seq, conditions)
                         for Q, seq in self.protein_sequences.items()}
        
        # Phase 2: TF partitioning
        partition_data = self.measure_TF_partitioning(
            condensates=phase_diagrams,
            TFs=self.transcription_factors
        )
        
        # Phase 3: Cellular validation
        cellular_data = {}
        for cell_line in self.cell_lines:
            frap = self.analyze_FRAP(measure_FRAP(cell_line))
            tf_nuclear = measure_nuclear_TF(cell_line)
            rnaseq = self.run_differential_expression(
                get_counts(cell_line), get_metadata(cell_line)
            )
            cellular_data[cell_line] = {'frap': frap, 'tf': tf_nuclear, 'rnaseq': rnaseq}
        
        # Phase 4: Compound screen
        screen_results = self.analyze_compound_screen(
            run_HCS(self.compound_library, 'STHdh_Q111'),
            controls=get_controls()
        )
        top_compounds = screen_results[:20]
        
        # Phase 5: In vivo
        mouse_results = run_mouse_study(top_compounds[:3])
        behavior = self.analyze_mouse_behavior(mouse_results['rotarod'],
                                                mouse_results['groups'])