Tumor-suppressors have commanded attention due to the selection for their inactivating mutations in human tumors. However, relatively little is understood about the inverse, namely, that tumors do not select for a large proportion of seemingly favorable mutations in tumor-suppressor genes. This could be explained by a detrimental phenotype accruing in a cell type-specific manner to most cells experiencing a biallelic loss. For example, MKK4, a tumor suppressor gene distinguished by a remarkably consistent mutational rate across diverse tumor types and an unusually high rate of loss of heterozygosity, has the surprisingly low rate of genetic inactivation of only approximately 5%. To explore this incongruity, we engineered a somatic gene knockout of MKK4 in human cancer cells. Although the null cells resembled the wild-type cells regarding in vitro viability and proliferation in plastic dishes, there was a marked difference in a more relevant in vivo model of experimental metastasis and tumorigenesis. MKK4(-/-) clones injected i.v. produced fewer lung metastases than syngeneic MKK4-competent cells (P = 0.0034). These findings show how cell type-specific detrimental phenotypes can offer a paradoxical and yet key counterweight to the selective advantage attained by cells as they experiment with genetic null states during tumorigenesis, the resultant balance then determining the observed biallelic mutation rate for a given tumor-suppressor gene.