Abstract
Detecting genetic variants enables risk factor identification, disease screening, and initiation of preventative therapeutics. However, current methods, relying on hybridization or sequencing, are unsuitable for point-of-care settings. In contrast, CRISPR-based-diagnostics offer high sensitivity and specificity for point-of-care applications. While these methods have predominantly been used for pathogen sensing, their utilization for genotyping is limited. Here, we report a multiplexed CRISPR-based genotyping assay using LwaCas13a, PsmCas13b, and LbaCas12a, enabling the simultaneous detection of six genotypes. We applied this assay to identify genetic variants in the APOL1 gene prevalent among African Americans, which are associated with an 8-30-fold increase in the risk of developing kidney disease. Machine learning facilitated robust analysis across a multicenter clinical cohort of more than 100 patients, accurately identifying their genotypes. In addition, we optimized the readout using a multi-analyte lateral-flow assay demonstrating the ability for simplified genotype determination of clinical samples. Our CRISPR-based genotyping assay enables cost-effective point-of-care genetic variant detection due to its simplicity, versatility, and fast readout.
Introduction
Genetic variants in the APOL1 gene, specifically the G1 and G2 alleles, are major risk factors for chronic kidney disease in populations of African ancestry. Individuals carrying two risk alleles face an 8- to 30-fold increased risk of disease. In clinical scenarios such as organ transplantation, rapid identification of these variants is critical for risk stratification and donor selection.
However, existing diagnostic standards like Sanger sequencing or qPCR are poorly suited for urgent clinical needs. These methods are often centralized, expensive, and suffer from long turnaround times—sometimes up to three weeks. While CRISPR-based diagnostics have revolutionized rapid pathogen detection, their application to human genotyping is technically demanding. Distinguishing single-nucleotide polymorphisms (SNPs) or small deletions requires extreme specificity to differentiate between heterozygous and homozygous states, a hurdle this research aims to overcome.
Methodology
The researchers developed a "one-pot" multiplexed assay capable of detecting wild-type (G0) and risk variants (G1, G2) simultaneously. The system integrates several key technical innovations:
Orthogonal CRISPR Detection
The assay utilizes three distinct Cas enzymes—LwaCas13a, PsmCas13b, and LbaCas12a—in a single reaction. By leveraging the unique "collateral" cleavage preferences of each enzyme, the team could use three different reporters to signal the presence of specific alleles.
Isothermal Amplification and Specificity Engineering
To facilitate point-of-care use, the assay employs Recombinase Polymerase Amplification (RPA), which operates at a constant 39°C, removing the need for thermal cyclers. To achieve the high specificity required for SNP detection (specifically the G1 variant), the team engineered "synthetic mismatches" into the CRISPR RNA (crRNA). This ensures the enzyme only triggers a signal when a perfect match is present.
Machine Learning and Lateral-Flow Readout
A Canonical Interval Forest machine learning model was developed to analyze fluorescence kinetics. By focusing on the shape of the signal curve over time, the model overcomes variability caused by different operators or equipment. For resource-limited settings, the team also optimized a multi-analyte lateral-flow "dipstick" assay that provides a visual pattern unique to each of the six possible APOL1 genotypes.
Clinical Validation
The assay was validated using genomic DNA from a multicenter cohort of 124 patients across the United States, Germany, and Brazil. Results were benchmarked against gold-standard Sanger sequencing and TaqMan qPCR.
Results
The multiplexed CRISPR assay demonstrated high performance across all metrics:
- Accuracy: The ML-assisted assay achieved 100% accuracy in identifying genotypes within the clinical cohort. Even using simple manual thresholding, the assay maintained a high accuracy of 97.9%.
- Speed: The total time from DNA extraction to result is approximately 2.5 to 3 hours, a significant improvement over traditional multi-day sequencing.
- Versatility: The assay successfully identified all six possible genotypes (G0/G0, G0/G1, G1/G1, G0/G2, G2/G2, G1/G2) in a single workflow.
- Point-of-Care Readiness: The lateral-flow version successfully identified genotypes visually, proving the method's viability in settings without fluorescence readers.
Discussion and Future Directions
This study demonstrates that combining multiple CRISPR enzymes with machine learning can create a robust, rapid, and cost-effective genotyping platform. The "Power of Three" approach allows for complex genetic analysis in a simplified workflow.
However, some limitations remain. The current protocol requires a 30-minute DNA extraction step; future work will focus on "sample-to-result" workflows using crude blood lysates. Additionally, the requirement for specific PAM sequences near target sites may limit the application to certain mutations. Moving forward, the team aims to transition this technology into clinical trials, specifically for the rapid screening of deceased organ donors to improve transplantation outcomes.
Summary updated