Since autumn 2020, clade 2.3.4.4b H5Nx high pathogenicity avian influenza virus (HPAIV) has undergone global spread, decimated poultry and resulting in significant losses in wild bird populations. Concerningly, there has been an increase in mammalian infections, which can promote adaptive mutations in these viruses which further increase their zoonotic risk. A well-known adaptive mutation, shown to increase viral replication in mammalian cells, is the substitution of glutamate (E) with lysine (K) at position 627 of the PB2 protein (PB2-E627K). PB2-E627K has been detected in clade 2.3.4.4b H5N1 HPAIV infecting mammals, including humans, during the epizootic, but not frequently detected in viruses from birds.   Here we investigated if mammalian-adaptive mutations could be maintained in birds. We investigated the impact of the PB2-E627K substitution on pathogenicity, transmissibility and stability in avian hosts in the context of a dominant European genotype (AB) of 2.3.4.4b H5N1 HPAIV.  In-vitro analysis showed that presence of PB2-627K significantly increased the replication efficiency of HPAIV in human cells compared to PB2-627E but does not significantly change the replication efficacy in duck or chicken cells. Similarly, PB2-627E or PB2-627K showed no significant difference in the infectivity, transmissibility, or pathogenicity in ducks or chickens, with the substitution remaining stable following transmission. The PB2-627K mutation is stable, has no fitness cost in avian hosts, yet it poses a significant zoonotic threat. This highlights the importance of continued surveillance in bird populations, and suggests mammalian-adapted variants can be maintained in wild birds, spill-over into poultry, and thus pose an elevated zoonotic risk.