The urgency behind the new assays is difficult to overstate. Since its first detection in January 2022, highly pathogenic avian influenza H5 has infected more than 13,000 wild birds and roughly 173 million poultry, with economic losses estimated between 14 and 164 billion dollars. In March 2024, an unprecedented spillover of clade 2.3.4.4b A(H5N1) into dairy cattle was reported in Texas, and the virus has since spread across 17 US states, infecting more than 1,000 cattle herds with sustained mammalian transmission and 70 human cases reported as of June 3, 2025. Historically, HPAI outbreaks such as H5N1 in Hong Kong in 1997 and H7N9 in China in 2013 have produced case fatality rates of at least 30 percent in humans, underscoring the pandemic potential that hangs over every new host adaptation.

Current diagnostic approaches each carry trade-offs that limit their usefulness in the field. Virus isolation remains the gold standard but is confined to biosafety level-3 laboratories staffed by highly trained personnel. Rapid immunoassays deliver results in about 15 minutes but suffer from reduced sensitivity and cannot discriminate between viral clades. Reverse transcription PCR offers high sensitivity and adaptability yet depends on thermocycling equipment and skilled operators, while next-generation sequencing provides nucleotide-level resolution at the cost of high expense and long turnaround times. The result, the authors argue, is a diagnostic landscape in which the tools that respond fastest to outbreaks are either too insensitive or too infrastructure-dependent to support surveillance at farms, markets, and rural clinics where spillovers first emerge.

To close this gap, the team turned to SHINE, short for Streamlined Highlighting of Infections to Navigate Epidemics, a one-pot CRISPR diagnostic platform previously developed for SARS-CoV-2 and influenza detection. In the avian influenza version of the workflow, viral RNA undergoes reverse transcription, recombinase polymerase amplification, and T7 transcription within a single tube, generating RNA amplicons that activate the collateral cleavage activity of the Cas13a enzyme when they base-pair with a complementary CRISPR RNA guide. Activated Cas13a then cuts RNA reporters, producing either a fluorescent signal read out on a plate reader or a colorimetric band on a paper lateral flow strip, the latter requiring nothing more sophisticated than a smartphone camera to document.

Assay design leaned heavily on machine learning. The researchers used ADAPT, a software platform that predicts sensitive and specific Cas13 guide sequences, to select primer and crRNA sets targeting the hemagglutinin gene segment that uniquely defines H5 viruses. Two candidate designs achieved predicted coverage of 98.72 and 96.18 percent of aligned H5N1 sequences, and the top-performing guide, targeting nucleotides 847 to 874 of the H5 consensus, was carried forward. A series of optimization experiments followed: the team tuned the ratio of T7-attached to non-T7-attached forward primers, settled on a 1:3 ratio that reduces competition between amplification and detection, increased RPA primer concentrations, and adjusted reporter, RNase H, and magnesium levels. The resulting optimized assay, SHINE-H5, reliably detected synthetic RNA targets across all tested concentrations within 60 minutes, a substantial improvement over the prototype.

Analytical characterization of SHINE-H5 was rigorous. Testing against six vaccine-derived viral seedstocks from the CDC showed high fluorescence for the intended target and negligible signal for non-target viruses. Clinical specificity was evaluated using 64 nasopharyngeal swab specimens from Massachusetts General Hospital that tested positive for seasonal influenza A or B but negative for avian H5 by qPCR; SHINE-H5 correctly returned negative results for all 64 samples, covering infections with H1N1, H3N2, and influenza B, including co-infections. The limit of detection, established with serial dilutions of H5N1 viral seedstock in viral transport media and a logistic regression model requiring at least 95 percent detection, was 121.7 copies per microliter, with a 95 percent confidence interval of 63.14 to 234.55 copies per microliter. In a head-to-head comparison, a validated RT-qPCR assay detected samples down to 10 copies per microliter while SHINE-H5 detected down to 50 copies per microliter, a sensitivity gap the authors acknowledge but one that still significantly outperforms existing H5 immunoassays.

For field deployment, the researchers adapted SHINE-H5 to a lateral flow readout, replacing the fluorescent quenched reporter with a FAM-biotin reporter whose cleavage products migrate on a paper strip. Using a larger 40-microliter reaction volume, the lateral flow version detected H5N1 seedstock down to 25 copies per microliter while retaining strong specificity against a subset of 14 seasonal influenza-positive patient samples. The team also demonstrated that the assay can detect synthetic RNA targets spiked into milk, a proof of concept for surveillance in the dairy herds now at the center of the North American outbreak.

Beyond broad H5 detection, the study introduces SHINE-H5-CS, a clade-specific assay targeting 2.3.4.4b A(H5N1), the lineage that has dominated global H5 phylogeny since 2021 and is driving the US cattle outbreak. Guide design for this assay employed BADGERS, a successor to ADAPT that integrates advanced search algorithms to explore the fitness landscape of candidate crRNAs and achieve single-nucleotide discrimination. Of three candidates, crRNA3, targeting a region of the H5 segment starting at nucleotide position 880, showed the highest on-target activity with minimal off-target signal. The researchers also shifted the forward primer four nucleotides upstream to introduce an additional mismatch that sharpened discrimination. SHINE-H5-CS consistently detected clade 2.3.4.4b RNA at concentrations above 100 copies per microliter while showing minimal cross-reactivity with non-2.3.4.4b H5 sequences and seasonal influenza seedstocks, enabling lineage identification without sequencing.

The team extended the platform to the H7 subtype, the other hemagglutinin group with well-documented potential to evolve into highly pathogenic strains. Phylogenetic analysis of 25 representative H7 strains revealed clean separation into Eurasian and North American lineages that mirror migratory bird flyways, and the 2013 Eurasian H7N9 epidemic, which caused more than 1,500 human infections with a fatality rate near 40 percent, demonstrated the public health stakes of this lineage. The resulting SHINE-H7-Eurasian assay reliably detected Eurasian H7 RNA across a range of concentrations while discriminating against North American H7 strains and a panel of unrelated seasonal influenza viruses, and it too was adapted to a lateral flow format that preserved lineage specificity.

The authors are candid about the limitations of the work. Confirmed human H5 and Eurasian H7 cases remain rare, so the assays could not be evaluated on true positive human clinical specimens, and compatibility with specimen types beyond mock milk samples, such as poultry or environmental samples, has not yet been assessed. Even so, the combination of isothermal operation at 37 degrees Celsius, turnaround times of 60 to 120 minutes, minimal instrumentation, machine-learning-accelerated assay design, and compatibility with paper-based readouts positions the SHINE family of assays as a practical bridge between laboratory-grade accuracy and point-of-need accessibility. As avian influenza continues its advance through wild birds, poultry, cattle, and occasionally people, tools that can detect not just the virus but its specific clades, close to where it emerges, may prove decisive in blunting the next spillover before it becomes the next pandemic.

Subject of Research: Development of streamlined one-pot CRISPR-Cas13 diagnostic assays for detection and genetic subtyping of H5 and H7 avian influenza viruses

Article Title: Streamlined CRISPR-based assays for detection and subtyping of H5 and H7 avian influenza

Article References: Huang, Y., Guo, A., Adams, G., Lemieux, J. E., & Myhrvold, C. (2026). Streamlined CRISPR-based assays for detection and subtyping of H5 and H7 avian influenza. iScience, 29(10), Article 117556. https://doi.org/10.1016/j.isci.2026.117556

Image Credits: AI Generated

DOI: 10.1016/j.isci.2026.117556

Keywords: avian influenza, CRISPR diagnostics, SHINE assay, Cas13, H5N1, H7 avian influenza, clade 2.3.4.4b, lateral flow assay, recombinase polymerase amplification, point-of-need testing, dairy cattle outbreak, pandemic preparedness