Integration of Epidemiological and Genomic Data to Investigate H5N1 HPAI Outbreaks in Northern Italy in 2021-2022

Abstract

Between October 2021 and April 2022, 317 outbreaks caused by highly pathogenic avian influenza (HPAI) H5N1 viruses were notified in poultry farms in the northeastern Italian regions. The complete genomes of 214 strains were used to estimate the genetic network based on the similarity of the viruses. An exponential random graph model (ERGM) was used to assess the effect of 'at-risk contacts', 'same owners', 'in-bound/out-bound risk windows overlap', 'genetic differences', 'geographic distances', 'same species', and 'poultry company' on the probability of observing a link within the genetic network, which can be interpreted as the potential propagation of the epidemic via lateral spread or a common source of infection. The variables 'same poultry company' (Est. = 0.548, C.I. = [0.179; 0.918]) and 'risk windows overlap' (Est. = 0.339, C.I. = [0.309; 0.368]) were associated with a higher probability of link formation, while the 'genetic differences' (Est. = -0.563, C.I. = [-0.640; -0.486]) and 'geographic distances' (Est. = -0.058, C.I. = [-0.078; -0.038]) indicated a reduced probability. The integration of epidemiological data with genomic analyses allows us to monitor the epidemic evolution and helps to explain the dynamics of lateral spreads casting light on the potential diffusion routes. The 2021-2022 epidemic stresses the need to further strengthen the biosecurity measures, and to encourage the reorganization of the poultry production sector to minimize the impact of future epidemics.

Keywords: ERGM; H5N1; HPAI; Italy; contact tracing; epidemiological investigation; genetic network.

Figure 1

Notified Italian HPAI H5N1 outbreak distribution in poultry farms and non-poultry birds (22 in wild birds and 1 in captive birds) during the 2021–2022 epidemic.

Figure 2

The epidemic curve showing the number of weekly cases in poultry farms and non-poultry birds (22 in wild birds and one in captive birds).

Figure 3

Phylogenetic tree of the hemagglutinin (HA) gene segment and genotypes identified among the Italian H5N1 viruses. (a) Schematic representation of the gene composition of each genotype. Light grey bars: conserved segments; dark grey bars: gene segments acquired by reassortment events. (b) Phylogenetic tree of the HA gene segment colored according to seven different genetic clusters. C, AC, AD, AF: different genotypes characterizing the Italian viruses; n.a.: viruses not assigned to genotypes due to the absence of one or more gene segments; pink cluster: genetic group that includes most of the Italian viruses collected from October 2021 to April 2022; dark yellow sequences: viruses collected from wild birds.

Figure 4

Geographic distribution of the domestic outbreaks in a densely populated poultry area. Pink dots: cluster A outbreaks; grey dots: non-clustered outbreaks or outbreaks belonging to other minor genetic groups.

Figure 5

Graphical representation of the estimated genetic network (cluster A). Pink nodes: H5N1 HPAI virus isolated from individual domestic outbreaks belonging to the larger genetic group; grey nodes: estimated median vectors; links: highest genetic similarity between nodes.

Figure 6

Graphical representation of the predicted probability distributions according to the (a) geographic distances, (b) risk time windows overlap, and (c) viral genetic differences. The x-axis scale for (a,c) was limited to 50 km and 15 bases, respectively, to improve the graph’s readability.

Figure 7

Tie prediction statistics, including the ROC (left panel) and PR (right panel) curves. Solid lines: actual ROC/PR curves; dashed lines: baseline ROC/PR curves drawn for a random graph.