Disclaimer: Early release articles are not considered as final versions. Any changes will be reflected in the online version in the month the article is officially released.

Author affiliations: Stanford Health Care Clinical Virology Laboratory, Stanford, California, USA (M.K. Sahoo, I.E.A. Morante, U.C. Ohiri, D. Romero, B.A. Pinsky); Stanford University School of Medicine, Stanford (C. Huang, D. Solis, F. Yamamoto, B.A. Pinsky)

Highly pathogenic avian influenza (HPAI) viruses cause high mortality rates in wild birds and farmed poultry, and their potential for adaptation to humans remains a major pandemic threat (1). In 2021, a notable increase in avian influenza activity occurred, driven by the emergence of influenza A(H5N1) clade 2.3.4.4b viruses (2,3). Suspected mammal-to-mammal transmission during outbreaks among seals and minks led to concerns regarding the escalation of risk to humans (4–7). In 2024, the virus was identified in dairy cows in the United States (8), and exposure to infected cows has resulted in transmission to humans (9). To prepare for possible additional human cases, we developed an internally controlled, dual-target H5 subtyping quantitative reverse transcription PCR (qRT-PCR) based on primer–probe sequences from the World Health Organization (10). We evaluated this qRT-PCR by using synthesized nucleic acids, cultured avian and human virus isolates, and clinical upper respiratory specimens collected from patients with influenza A virus infection.

We identified previously published primer–probe sets from the World Health Organization that target different regions of the influenza A(H5) hemagglutinin (HA) gene for combination into a dual-target H5 subtyping qRT-PCR (Appendix Table 1) (10). We introduced minor sequence changes to account for recent clade 2.3.4.4b diversity, and simultaneously, to limit primer–probe complexity (Table 1). We evaluated these primers and probes against complete North American H5 clade 2.3.4.4b HA gene sequences in the GISAID database (https://www.gisaid.org) during January 1, 2022–May 29, 2024, and found that 99.8% (5,975/5,987) of sequences aligned with a maximum of 1 mismatch to primer–probe set 1 and 97.7% (5,972/5,990) aligned with a maximum of 1 mismatch to primer–probe set 2 (Appendix Tables 2, 3). We observed no primer–probe mismatches compared with the influenza A H5 sequences obtained from persons with H5 infections in the United States (4 persons as of May 29, 2024) (Appendix Tables 4, 5). We combined the primer–probe sets in multiplex with primers–probes for the influenza A matrix (M) gene for pan–influenza A detection and with primers–probes for RNase P as an internal control (11).

We determined the 95% lower limit of detection (LLOD) by using single-stranded DNA (ssDNA) encoding the H5 target sequences from HPAI virus clade 1 (GenBank accession no. JQ966928, A/condor/Guangdong/139/2003 [H5N1]) (12) and clade 2.3.4.4b (GenBank accession no. OP499866, A/red-tailed hawk/Kansas/W22–198/2022 [H5N1]) viruses (Appendix Table 6). The primer–probe target regions of this clade 2.3.4.4b H5 sequence are identical to the recent H5 sequences from dairy cows and humans. We combined the 2 clade 1 ssDNA targets at equal copy numbers; we did the same with the clade 2.3.4.4b ssDNA targets. We also combined these H5 clade mixes at equal copy numbers with ssDNA encoding the pan–influenza A M gene target. We then diluted the clade 1 ssDNA mix and clade 2.3.4.4b ssDNA mix to 2, 1, and 0.5 copies/μL in pooled influenza A–negative eluates extracted from clinical upper respiratory specimens. We then performed the H5 subtyping qRT-PCR (Appendix). We tested 20 replicates at each level for both mixes. The 95% LLOD for the H5 target was 2.5 copies/μL (95% CI 1.8–5.3 copies/μL) for the clade 1 ssDNA mix and <0.5 copies/μL (we were unable to calculate 95% CI) for the clade 2.3.4.4b ssDNA mix.

We also evaluated the H5 subtyping multiplex qRT-PCR by using genomic RNA from a clade EurAsian non-goose/GuangDong (Gs/Gd) isolate, Kilbourne F181, A/duck/Singapore/645/1997 (H5N3). We assigned this genomic RNA an estimated concentration in copies/μL by using the clade 2.3.4.4b ssDNA as standard curve, then diluted it to 2, 1, and 0.5 copies/μL in pooled influenza A–negative eluates. We tested 20 replicates at each level. The 95% LLOD for the H5 target was 0.6 copies/μL (95% CI 0.5–0.7 copies/μL). In addition, we tested 10-fold serial dilutions from 4.6 to 0.6 log10 copies/mL of the Kilborne F181 genomic RNA in duplicate (Appendix Figure), which indicated 0.97 reaction efficiency for the H5 target.

We then evaluated the H5 subtyping qRT-PCR by using genomic RNA from additional H5 viruses, including 5 low pathogenicity avian influenza non-Gs/Gd isolates from the United States (Table 2). We tested each eluate in duplicate. Although we detected H5 RNA in all reactions, the difference in mean cycle threshold (Ct) values between the H5 and pan–influenza A (M gene) targets varied widely (range 0.6–9.9 cycles). Testing the H5 primer–probe sets individually revealed that the US non-Gs/Gd genomic RNAs were detected only by primer–probe set 1. The A/mallard/Minnesota/16-041335-3/2016 (H5N2) eluate showed the greatest Ct difference, consistent with the highest number of H5 primer–probe mismatches in primer–probe set 1 (9 mismatches) (Appendix Tables 5, 6).

To determine specificity, we tested genomic RNA from cultured human and avian virus isolates encoding non-H5 HA genes (Table 3). We detected the pan–influenza A target in all eluates; median Ct value was 17.3 (range 13.6–20.7). We did not detect the H5 target in any eluates.

We performed further specificity experiments by using residual upper respiratory swab samples in viral transport media submitted for clinical respiratory virus testing. We conducted this work with Stanford Institutional Review Board approval (protocol no. 68234); individual consent was waived. We evaluated 100 samples collected during the 2022–2023 respiratory virus season that were positive for influenza A by using routine clinical methods (98 by using Hologic Panther Fusion [Hologic, https://www.hologic.com] and 2 by using Cepheid GeneXpert [Cepheid, https://www.cepheid.com]), including 29 known co-infections (13 respiratory syncytial virus [RSV], 7 rhinovirus, 3 adenovirus, 3 human metapneumovirus, 1 parainfluenza-3, 1 parainfluenza 4, and 1 with both RSV and rhinovirus). In addition, we tested 55 upper respiratory samples collected in April 2024 that were negative for influenza A, influenza B, RSV, parainfluenza virus 1–4, rhinovirus, adenovirus, and human metapneumovirus on the Panther Fusion by using the H5 subtyping qRT-PCR. We extracted total nucleic acids from 400 µL and eluted them in 60 µL buffer AVE on the BioRobot EZ1 (Qiagen) by using the EZ1 virus mini kit 2.0, according to the manufacturer’s recommendations. We detected the pan–influenza A target in 97.0% (97/100) of the influenza A–positive samples; median Ct value was 24.1 (range 12.6–37.7). All 3 of the samples that were not detected in the pan–influenza A channel of the H5 multiplex had detectable RNase P (Ct values of 22.5, 18.7, and 20.8), indicating adequate extraction and the absence of inhibitors. The original influenza A Ct values (Panther Fusion) for those 3 samples were 41.4, 38.0, and 38.2. Upon repeat testing, influenza A RNA was detected only in the sample with the lowest original Ct value for influenza A (Ct 34.8) by the H5 multiplex. None of those influenza A–positive samples had detectable H5 RNA. Of the respiratory virus panel negative samples, all had detectable RNase P; median Ct value was 23.9 (range 17.7–29.8). None of the panel negative samples had detectable H5 RNA.

HPAI H5 viruses, particularly clade 2.3.4.4b viruses, pose an emerging pandemic threat. This internally controlled dual-target influenza A(H5) qRT-PCR demonstrated high analytical performance and could be used to directly test samples from patients with suspected HPAI A(H5) virus infection or as a secondary test to subtype known influenza A–positive samples after routine respiratory virus testing. The dual-target design reduces the likelihood of false negatives, as evidenced by detection of low pathogenicity avian influenza H5 viruses with substantial target mismatch. However, continuous sequence surveillance and updating of the primer–probe sets will be required to ensure that this assay accounts for ongoing viral evolution (13).

Dr. Sahoo is senior bioinformatic staff scientist in the Stanford Health Care Clinical Virology Laboratory. His research involves the design and evaluation of nucleic acid amplification and sequencing assays for the diagnosis and characterization of infectious diseases.

Top