Human MxA protein is an interferon-induced 76-kDa GTPase that exhibits antiviral activity against several RNA viruses. Wild-type MxA accumulates in the cytoplasm of cells. TMxA, a modified form of wild-type MxA carrying a foreign nuclear localization signal, accumulates in the cell nucleus. Here we show that MxA protein is translocated into the nucleus together with TMxA when both proteins are expressed simultaneously in the same cell, demonstrating that MxA molecules form tight complexes in living cells. To define domains important for MxA-MxA interaction and antiviral function in vivo, we expressed mutant forms of MxA together with wild-type MxA or TMxA in appropriate cells and analyzed subcellular localization and interfering effects. An MxA deletion mutant, MxA(359-572), formed heterooligomers with TMxA and was translocated to the nucleus, indicating that the region between amino acid positions 359 and 572 contains an interaction domain which is critical for oligomerization of MxA proteins. Mutant T103A with threonine at position 103 replaced by alanine had lost both GTPase and antiviral activities. T103A exhibited a dominant-interfering effect on the antiviral activity of wild-type MxA rendering MxA-expressing cells susceptible to infection with influenza A virus, Thogoto virus, and vesicular stomatitis virus. To determine which sequences are critical for the dominant-negative effect of T103A, we expressed truncated forms of T103A together with wild-type protein. A C-terminal deletion mutant lacking the last 90 amino acids had lost interfering capacity, indicating that an intact C terminus was required. Surprisingly, a truncated version of MxA representing only the C-terminal half of the molecule exerted also a dominant-negative effect on wild-type function, demonstrating that sequences in the C-terminal moiety of MxA are necessary and sufficient for interference. However, all MxA mutants formed hetero-oligomers with TMxA and were translocated to the nucleus, indicating that physical interaction alone is not sufficient for disturbing wild-type function. We propose that dominant-negative mutants directly influence wild-type activity within hetero-oligomers or else compete with wild-type MxA for a cellular or viral target.