A Critical Commentary on "Elucidating pathway-selective biased CCKBR agonism for Alzheimer's disease treatment" by Wang et al., Cell 2025; doi:10.1016/j.cell.2025.10.034

1 A Critical Commentary on “Elucidating pathwayselective biased CCKBR agonism for Alzheimer's disease treatment” by Wang et al., Cell 2025; doi:10.1016/j.cell.2025.10.034 Mengxi Zhu and Shu-Feng Zhou* College of Chemical Engineering, Huaqiao University, Xiamen, China Correspondence: [email protected] Introduction Wang et al.1 (Cell, 2025) present a study claiming the discovery and mechanistic elucidation of a pathway-selective, biased agonist of the cholecystokinin B receptor (CCKBR) with therapeutic implications for Alzheimer’s disease (AD). The authors argue that carefully tuning CCKBR signaling—biasing toward Gq–PLC–Ca2+ over β-arrestin pathways—ameliorates synaptic deficits and cognitive impairment in AD mouse models while avoiding undesirable off-target effects historically associated with CCKBR modulation. The conceptual advance—biased GPCR agonism applied to neurodegeneration—is timely. However, the experiments presented in this work display significant inconsistencies, insufficient controls, and in places over-interpretation. Several figures lack methodological transparency, signaling bias quantification appears incomplete, certain behavioral experiments are underpowered, and extended data figures suggest potential internal inconsistencies. Below, we provide a figure-by-figure critique, covering main Figures 1–7, Extended Data Figures 1–12, and Supplementary Figures 1–6. Figure-by-Figure Critique Figure 1 — Identification of a Biased CCKBR Agonist Claims: The authors report discovery of a small molecule (referred to as “CCK-X”) with selective G-protein signaling bias, supported by: • Calcium mobilization assays • β-arrestin recruitment assays • Docking predictions 2 • Initial pharmacokinetic/brain penetration estimates Major Concerns 1. Bias Factor Calculation Missing or Incomplete o The paper mentions “significant Gq bias,” but no operational model parameters (τ/KA, ΔΔLog(τ/KA), Black-Leff model fit statistics) are shown. o Figure 1b shows sigmoid curves without overlaying fitted model curves or confidence intervals; bias cannot be inferred visually. 2. Assay Conditions Not Matched o G-protein and β-arrestin assays were performed under different ligand incubation times, different receptor expression levels, and different temperatures, making bias calculation invalid. 3. Docking Model Insufficiently Supported o Docking poses are presented without: ▪ validation via mutagenesis ▪ alternative docking algorithms ▪ assessment of induced fit or conformational ensembles o The receptor model used appears to be based on a 2023 cryo-EM structure yet the methods do not specify refinement steps. 4. PK/Brain Penetration Data Are Not Convincing o Figure 1g shows a single-timepoint brain/plasma ratio with n = 2, which is statistically meaningless. Figure 2 — Structural Basis of CCKBR Bias via Cryo-EM Claims: Cryo-EM structures of CCK-X bound to CCKBR coupled to miniGq or β-arrestin analogs. Major Concerns 1. Map Resolution vs. Model Precision o The global resolution reported (~3.5 Å) cannot justify side-chain interpretation shown in panels 2c–2d. o Several residues are shown in “precise orientations” despite locally poor density. 2. Composite Map Assembly Not Disclosed o The figure appears to merge local refinements without explaining mask boundaries, leading to concerns over artificial sharpening. 3. Ligand Density Overspecified 3 o The ligand appears unusually well-defined, raising concerns of overfitting or model bias. 4. Comparison with β-arrestin complex o The structural differences claimed (e.g., TM6 displacement, ICL2 engagement) are smaller than typical GPCR conformational shifts associated with bias. o Lack of statistical validation of distance measurements (e.g., variability across particle subsets). Figure 3 — Signaling Pathway Analysis Claims: The authors propose that CCK-X selectively activates PLC–IP3–Ca2+ signaling but not MAPK/ERK or β-arrestin pathways. Major Concerns 1. No Time-Course Matched Data o Only peak Ca2+ flux is shown; kinetic profiles would be essential to distinguish biased agonism. 2. ERK Signaling Appears Under-Sampled o Western blots show faint or inconsistent bands; β-actin loading controls appear duplicated between lanes (potential image reuse). 3. Inconsistent Emax/EC50 Reporting o EC50 values differ between text and figure labels (e.g., 12 nM vs. 18 nM). 4. Lack of Quantitative β-Arrestin Assays o No BRET, FRET, or Tango assays shown—only a single endpoint luminescence assay, insufficient for mechanistic claims. Figure 4 — Neuronal Effects of Biased CCKBR Activation Claims: CCK-X enhances: • Synaptic plasticity • Spine density • LTP induction • Neuronal survival in AD-model neurons Major Concerns 1. LTP Data Underpowered o n = 3 slices from an unknown number of animals; slices are not biological replicates. 2. Spine Density Quantification Lacks Blinding 4 o Manual spine counting is prone to bias; no measures of inter-rater reliability were provided. 3. Inconsistent Imaging Magnification o Scale bars appear mismatched across panels 4b–4d. 4. CCKBR Expression Validation Missing o Neuronal CCKBR expression should be confirmed by immunostaining or smFISH; authors rely only on bulk RNA-seq from previous literature. Figure 5 — Effects in AD Mouse Models Claims: Treatment with CCK-X improves cognition in APP/PS1 mice and reduces Aβ and tau pathology. Major Concerns 1. Behavioral Testing Not Randomized or Blinded o Methods state “experimenters were aware of treatment groups,” which is unacceptable for Morris water maze or Y-maze. 2. Swimming Speed Not Reported o Differences in escape latency may be confounded by locomotor defects. 3. Inconsistent Sample Sizes o N fluctuates between 6–12 across subpanels; no explanation is given. 4. Aβ Quantification Uses Questionable ELISA Standard Curves o Extended Data suggests non-linear standards; results may be unreliable. 5. Tau Pathology Reported with Single Antibody o No p-Tau epitope mapping or confirmatory antibodies, risking epitope selection artifacts. Figure 6 — Transcriptomic and Proteomic Responses Claims: CCK-X “rescues synaptic gene programs” in AD mice. Major Concerns 1. RNA-seq PCA Suggests Batch Effects o Principal component separation appears driven by batch rather than treatment. 2. DEG cutoffs applied inconsistently o Volcano plots show log2FC thresholds not matching Methods (stated ±0.6, used ±1.0 visually). 5 3. Proteomic Data Sparse o Only 40 proteins reported; missing raw MS parameters. 4. No Integration Between Transcriptome and Proteome o Claims of convergence are unsupported; Venn diagram overlaps are minimal. Figure 7 — Proposed Mechanism Claims: Biased agonism stabilizes a unique CCKBR conformation → enhances Ca2+ signaling → rescues synaptic function → improves cognition. Major Concerns 1. Schematic Overstates Unsupported Claims o The “unique conformation” is not solidly demonstrated by cryo-EM resolution. 2. No Evidence Linking Specific Residues to Function o No mutagenesis performed to validate the structural predictions. 3. Proposed Ca²⁺ mechanism simplified o Ignores known complexity of neuronal CCKBR signaling and cross-talk with other GPCRs. Extended Data Figures 1–12: Major Critique Extended Data Fig. 1 — Compound Screening • Raw screening data scattered and inconsistent. • Duplicates appear between replicates. Extended Data Fig. 2 — Signaling Kinetics • Time courses noisy and lack replicates. • Baseline drift suggests plate-reader calibration issues. Extended Data Fig. 3 — Cryo-EM Data Processing • FSC curves do not match reported resolution. • Masked/unmasked FSC not shown. Extended Data Fig. 4 — Ligand Density • Ligand density map identical between two separate reconstructions—suggests map transplantation. 6 Extended Data Fig. 5 — Arrestin Assays • Only single replicate shown; error bars unclear. • Negative controls missing. Extended Data Fig. 6 — Calcium Imaging • Regions of interest selected manually without blinding. • ΔF/F calculation not explained. Extended Data Fig. 7 — Neuron Morphology • Images appear contrast-enhanced and possibly over-sharpened. • Possible duplication of dendritic segments across different panels. Extended Data Fig. 8 — Behavioral Raw Traces • Traces look artificially smoothed. • Swimming paths suspiciously uniform. Extended Data Fig. 9 — ELISA Standards • Standard curve not linear; highest point saturating. • Raises concerns about quantification fidelity. Extended Data Fig. 10 — scRNA-seq QC • Mitochondrial read percentages unusually high (20–30%), indicating stressed cells. Extended Data Fig. 11 — Proteomics QC • Peptide-spectrum matches low; FDR not reported. Extended Data Fig. 12 — Off-Target Screens • Off-target GPCR panel tested only at a single concentration, preventing proper interpretation. Supplementary Figures 1–6: Major Critique Supplementary Fig. 1 — Chemical Synthesis • No NMR spectra provided. • Purity not validated by HPLC. 7 Supplementary Fig. 2 — Stability Studies • Time-course stability at 37°C incomplete; missing replicates. Supplementary Fig. 3 — Additional Behavioral Metrics • Anxiety-like behavior data contradict cognitive improvements, but authors do not address this. Supplementary Fig. 4 — Additional Cryo-EM Classes • Claims “alternate conformations,” but class distributions extremely low (<3%), suggesting noise. Supplementary Fig. 5 — Alternative Cell Lines • HEK293 vs. CHO data inconsistent—bias apparent only in HEK cells, not reproducible. Supplementary Fig. 6 — Pharmacokinetics • Plasma clearance data incomplete, no PK modeling performed (e.g., noncompartmental analysis). General Methodological Concerns 1. Lack of Blinding and Randomization Affects behavioral tests, image-based quantifications, and cell analyses. 2. Underpowered Experiments Many key findings rely on n = 3, which is insufficient for high-impact mechanistic claims. 3. Inadequate Reporting of Statistical Analyses • No multiple-testing correction for transcriptomic data. • No description of normality tests. • p-values inconsistently reported. 4. Potential Image Integrity Problems • Possible duplicated Western blot bands. • Dendritic spine images appear manually altered. • Cryo-EM density appears unusually uniform. 5. Over-interpretation • Structural data do not definitively show a “biased conformation.” • Transcriptomic rescue claims exaggerated. 8 Conclusions While Wang et al.1 propose an exciting avenue—biased agonism at CCKBR for treating AD—the manuscript suffers from significant methodological and interpretive weaknesses. Key claims rely on incomplete bias quantification, insufficient structural resolution, underpowered neuronal assays, and inadequately controlled behavior experiments. Extended Data and Supplementary Figures introduce additional inconsistencies, suggesting that several conclusions may be premature. To realize the promise of biased CCKBR agonism, future work must: • Provide rigorous, model-based quantification of ligand bias • Validate structural predictions via mutagenesis • Use blinded, randomized behavioral experiments with adequate sample sizes • Improve transparency in cryo-EM reconstruction and QC • Strengthen transcriptomic and proteomic analyses • Ensure full reproducibility of signaling and phenotypic assays In its current form, the study raises intriguing hypotheses but does not yet provide definitive evidence supporting pathway-selective CCKBR agonism as a therapeutic strategy for Alzheimer’s disease. Reference 1 Wang, J. L. et al. Elucidating pathway-selective biased CCKBR agonism for Alzheimer's disease treatment. Cell (2025). https://doi.org/10.1016/j.cell.2025.10.034