ABSTRACT Cleavage of the hemagglutinin (HA) by host proteases is essential for the infectivity of influenza viruses. Here, we analyzed the role of the serine protease TMPRSS2, which activates HA in the human respiratory tract, in pathogenesis in a mouse model. Replication of the human H7N9 isolate A/Anhui/1/13 and of human H1N1 and H3N2 viruses was compared in TMPRSS2 knockout ( TMPRSS2 ^−/− ) and wild-type (WT) mice. Knockout of TMPRSS2 expression inhibited H7N9 influenza virus replication in explants of murine tracheas, bronchi, and lungs. H1N1 virus replication was also strongly suppressed in airway explants of TMPRSS2 ^−/− mice, while H3N2 virus replication was only marginally affected. H7N9 and H1N1 viruses were apathogenic in TMPRSS2 ^−/− mice, whereas WT mice developed severe disease with mortality rates of 100% and 20%, respectively. In contrast, all H3N2 infected TMPRSS2 ^−/− and WT mice succumbed to lethal infection. Cleavage analysis showed that H7 and H1 are efficiently activated by TMPRSS2, whereas H3 is less susceptible to the protease. Our data demonstrate that TMPRSS2 is a host factor that is essential for pneumotropism and pathogenicity of H7N9 and H1N1 influenza virus in mice. In contrast, replication of H3N2 virus appears to depend on another, not yet identified protease, supporting the concept that human influenza viruses differ in protease specificity. IMPORTANCE Cleavage of the hemagglutinin (HA) by host proteases is essential for the infectivity of influenza virus, but little is known about its relevance for pathogenesis in mammals. Here, we show that knockout mice that do not express the HA-activating protease TMPRSS2 are resistant to pulmonary disease with lethal outcome when infected with influenza A viruses of subtypes H7N9 and H1N1, whereas they are not protected from lethal H3N2 virus infection. These findings demonstrate that human influenza viruses differ in protease specificity, and that expression of the appropriate protease in respiratory tissues is essential for pneumotropism and pathogenicity. Our observations also demonstrate that HA-activating proteases and in particular TMPRSS2 are promising targets for influenza therapy.