Emergence and genetic variation of neuraminidase stalk deletions in avian influenza viruses

Abstract

When avian influenza viruses (AIVs) are transmitted from their reservoir hosts (wild waterfowl and shorebirds) to domestic bird species, they undergo genetic changes that have been linked to higher virulence and broader host range. Common genetic AIV modifications in viral proteins of poultry isolates are deletions in the stalk region of the neuraminidase (NA) and additions of glycosylation sites on the hemagglutinin (HA). Even though these NA deletion mutations occur in several AIV subtypes, they have not been analyzed comprehensively. In this study, 4,920 NA nucleotide sequences, 5,596 HA nucleotide and 4,702 HA amino acid sequences were analyzed to elucidate the widespread emergence of NA stalk deletions in gallinaceous hosts, the genetic polymorphism of the deletion patterns and association between the stalk deletions in NA and amino acid variants in HA. Forty-seven different NA stalk deletion patterns were identified in six NA subtypes, N1-N3 and N5-N7. An analysis that controlled for phylogenetic dependence due to shared ancestry showed that NA stalk deletions are statistically correlated with gallinaceous hosts and certain amino acid features on the HA protein. Those HA features included five glycosylation sites, one insertion and one deletion. The correlations between NA stalk deletions and HA features are HA-NA-subtype-specific. Our results demonstrate that stalk deletions in the NA proteins of AIV are relatively common. Understanding the NA stalk deletion and related HA features may be important for vaccine and drug development and could be useful in establishing effective early detection and warning systems for the poultry industry.

Figure 1. Patterns and prevalence of AIV SΔNA.

Bars represent positions of missing amino acid residues. Colors indicate NA subtypes (red  =  N1, blue  =  N2, orange  =  N3, green  =  N5, purple  =  N6 and brown  =  N7). The first column on the left shows the deletion IDs assigned to each pattern. The second column shows the number of sequences having the corresponding deletion pattern and in parentheses the percentage of this pattern among all sequences of the same NA subtype. Percentages do not sum to 100 for the entire graph since they are calculated per NA subtype.

Figure 2. Phylogenetic tree of a subset of N1 sequences and distribution of SΔNA patterns (constructed in MrBayes 3.1.2).

Values at nodes show estimated posterior probabilities for bipartitions. Branch colors indicate host order. Genes of isolates from gallinaceous hosts are shown by tree branches with extended grey lines. The first column from the left shows a black dash for each isolate with SΔNA. The second column shows the deletion pattern ID. Sequences with the same deletion pattern are denoted by square brackets numbered with pattern ID. The third column shows which continent each isolate was from and the fourth column indicates the HA subtype of each isolate.

Figure 3. Spatial-temporal distribution of SΔNA patterns.

Each line indicates an instance at which an SΔNA pattern emerged (refer to Methods for definition). Pattern IDs on the left are the same as in Figure 1. Deletion patterns that are distributed over several distinct clades are indicated by lower case letters. Each cross indicates a year at which a particular pattern was observed. Crosses are color-coded for NA subtype. Bars on the right indicate the continent in which the deletion pattern was most commonly observed. Dotted time lines indicate deletion patterns that are not shown in the phylogenetic trees because their NA genes were partially sequenced.

Figure 4. Proportions of HA protein features among viruses with full-length-NA (filled bars) or with SΔNA (open bars).

The X-axis indicates the features, where Glyc denotes a glycosylation site, Del denotes a deletion, Ins denotes an insertion and the number indicates the amino acid position according to the H3 numbering. NA subtypes are color-coded (red  =  N1, blue  =  N2, orange  =  N3, brown  =  N7). Asterisks denote HA features whose SΔNA are statistically correlated with the HA feature since the posterior probability that ΔQ exceeds zero is greater than 0.95 (refer to Methods and Table 4).

Figure 5. Illustration of all possible combinations of two binary characters (A and B) and the transition rates between these combinations.

The model fits rates for the transitions per infinitely small time interval and therefore allows the change of only one state at a time (i.e. no rates are fitted for diagonal transitions). The shaded character combinations are expected to dominate if characters A and B are positively correlated. This positive correlation is created by transition rates if ΔQ = q ₂₁+q ₂₃+q ₄₁+q ₄₃−(q ₁₂+q ₁₄+q ₃₂+q ₃₄)>0.