Integrity Review Report: Cas9 gene therapy for Angelman syndrome traps Ube3a-ATS long non-coding RNA (DOI: 10.1038/s41586-020-2835-2)

Academic fraud report · Geng Detector

Summary

Verdict: No evidence of academic misconduct detected. This Nature 2020 paper by Wolter, Mao, Zylka and colleagues reports an AAV-CRISPR approach to Angelman syndrome. The text-based forensic review across six standard integrity dimensions found no red flags. Key observations supporting a clean finding include: (1) reporting of incomplete and variable UBE3A protein rescue across tissues (Cortex ~37%, Hippocampus ~38%, Cerebellum ~40%, no significant change in spinal cord), with congruent partial behavioral rescue; (2) transparent reporting of null results in behavioral assays and non-significant p-values (e.g., P = 0.09); (3) methodologically appropriate statistics (two-tailed unpaired t-test, one-way ANOVA, Kruskal–Wallis) with declared blinding and a priori power calculations; (4) a credible timeline consistent with 2017–2019 experiments including 17-month-old mice and 10-month safety monitoring; (5) intellectual honesty in revising the mechanistic hypothesis after sequencing revealed only 0.3% indels, leading to the discovery of AAV-integration-mediated transcriptional collision. Limitations: the analysis relied on text and figure legends only; high-resolution image-level forensics (pixel duplication, splicing artifacts, noise inconsistencies) were not performed.

Verdict

Cleared (no evidence of misconduct). Text-based integrity audit across six standard dimensions found no anomalies. Image-level pixel forensics were not performed due to lack of high-resolution source files.

Key findings

No image reuse / control logic issues: Western blot comparisons in Fig. 2c/d across WT, AS, and AS+Sajw33 groups show biologically plausible, non-uniform rescue (Cortex ~37%, Hippocampus ~38%, Cerebellum ~40%, Spinal cord not significantly rescued), consistent with the partial behavioral rescue reported in Fig. 3.
No data fabrication indicators: Behavioral data include explicitly reported null results (no rescue in distance travelled or marble burying; no deficits in contextual-based learning in AS mice in the authors' hands) and non-significant p-values (e.g., P = 0.09).
No statistical irregularities: Methods specify two-tailed unpaired Student's t-test, one-way ANOVA, and Kruskal–Wallis test as appropriate; Reporting Summary confirms blinding ("Investigators were blind to genotype and treatment") and prior power-based sample size calculation.
No timeline / reagent inconsistencies: SaCas9 (Ran et al., 2015) and lentiCRISPR v2 (Sanjana et al., 2014) were both available well before the Oct 2019 submission; 17-month expression and 10-month safety windows are compatible with experiments initiated ~2017–2018.
No methodological contradictions: The authors transparently revise their initial CRISPR-cut mechanism after observing only ~0.3% indels by sequencing, redirecting to AAV-integration-mediated transcriptional collision (Fig. 4). Animal ethics approval (18-183) is documented.

Evidence highlights

Quantitative UBE3A protein rescue values preserved exactly as reported: Cortex ~37%, Hippocampus ~38%, Cerebellum ~40%; Spinal cord not significantly changed.
Reported indel frequency of 0.3% at the on-target site, contradicting the original cut-and-degrade hypothesis and prompting the AAV-collision model.
Transparent acknowledgement of experimental artefacts: Fig. 4a notes that PCR "resulted in a smear (not shown) due to the repetitive binding site," consistent with the primer-walking strategy described.
DOI: 10.1038/s41586-020-2835-2; Received 11 October 2019; Accepted 28 July 2020.

Notes

Limitations: This review is text- and figure-legend-based only. Pixel-level duplicate-image detection, noise-consistency analysis, and splicing-edge checks were not conducted because original high-resolution images and raw gels/blots were unavailable. A residual possibility of image manipulation cannot be ruled out without direct image forensics.
Scope: Findings are limited to textual logic, statistical reporting, timeline plausibility, and methodological self-consistency. Institutional investigation would be required for a definitive determination, though no basis for referral was identified.
Recommendation: No action (PubPeer post, author contact, or editorial notification) warranted on the basis of this review.