Two surface glycoproteins of influenza virus, haemagglutinin (HA) and neuraminidase (NA), play opposite roles in terms of their interaction with host sialic acid receptors. HA attaches to sialic acid on host cell surface receptors to initiate virus infection while NA removes these sialic acids to facilitate release of progeny virions. This functional opposition requires a balance. To explore what might happen when NA of an influenza virus was replaced by one from another isolate or subtype, in this study, we generated three recombinant influenza A viruses in the background of A/PR/8/34 (PR8) (H1N1) and with NA genes obtained respectively from the 2009 pandemic H1N1 virus, a highly pathogenic avian H5N1 virus, and a lowly pathogenic avian H9N2 virus. These recombinant viruses, rPR8-H1N1NA, rPR8-H5N1NA, and rPR8-H9N2NA, were shown to have similar growth kinetics in cells and pathogenicity in mice. However, much more rPR8-H5N1NA and PR8-wt virions were released from chicken erythrocytes than virions of rPR8-H1N1NA and rPR8-H9N2NA after 1 h. In addition, in MDCK cells, rPR8-H5N1NA and rPR8-H9N2NA infected a higher percentage of cells, and induced cell-cell fusion faster and more extensively than PR8-wt and rPR8-H1N1NA did in the early phase of infection. In conclusion, NA replacement in this study did not affect virus replication kinetics but had different effects on infection initiation, virus release and fusion of infected cells. These phenomena might be partially due to NA proteins' different specificity to α2-3/2-6-sialylated carbohydrate chains, but the exact mechanism remains to be explored.