Abstract: FR-PO269
Deep Learning Allows Automated Analysis of Cystogenesis in Novel Xenopus Animal Models for Autosomal Polycystic Kidney Disease
Session Information
- Genetic Diseases of the Kidneys: Cystic - II
November 04, 2022 | Location: Exhibit Hall, Orange County Convention Center‚ West Building
Abstract Time: 10:00 AM - 12:00 PM
Category: Genetic Diseases of the Kidneys
- 1101 Genetic Diseases of the Kidneys: Cystic
Authors
- Naert, Thomas, Universitat Zurich, Zurich, ZH, Switzerland
- Lienkamp, Soeren S., Universitat Zurich, Zurich, ZH, Switzerland
Background
Autosomal dominant polycystic kidney disease (ADPKD) is caused by mutations in PKD1 or PKD2 and has an unmet need for new drugs and therapeutic targets. Ideally, these are identified in clinically relevant vertebrate disease models amenable to screening efforts. Further, genome editing simplifies the generation of new animal models for congenital disorders. However, the detailed and unbiased phenotypic assessment of altered embryonic development remains a challenge.
Methods
We employ targeted and unilateral CRISPR/Cas9 editing in order to inactivate pkd1 or pkd2 in the Xenopus developing vertebrate kidney. Cystogenesis is then visualized by whole-mount immunostaining, fluorescent stereomicroscopy, mesoSPIM light-sheet microscopy and advanced U-Net deep learning image processing for automated, unbiased and rapid scoring of kidney pathological states in two and three dimensions.
Results
CRISPR/Cas9 genome engineering in pkd1 and pkd2 elicited cystic malformations in developing renal tubules two-days post-fertilization (p<0.001). We observed cystogenesis across different developmental stages by leveraging an image processing pipeline for automated scoring of ADPKD in Xenopus embryos using deep learning approaches. Using a combination of segmentation and classification deep learning architectures allowed for automated size measurement of kidneys, as well as a qualitative analysis of cystic hallmarks. Our models correlated well with an independent expert on test data (ntest=120; r=0.96; P<0.001). Next, using tissue clearing and light-sheet microscopy approaches, we extended to three-dimensional analysis of cystogenesis. Here we showed that three-dimensional kidney size quantification can be achieved.
Conclusion
By combining light-sheet microscopy and deep learning we provide a framework for higher-throughput and in-depth characterization of novel models for autosomal polycystic kidney disease.
MesoSPIM light-sheet microscopy allows for whole-embryo imaging of Xenopus tropicalis embryos. Wild-type (top) embryos develop normal kidneys, pkd1 mutants (bottom) reveal cystogenesis.