HIV is famous for dodging the immune system, but that elusiveness doesn't come cheap. As Crawford et al. reveal on page 909, mutations that help to conceal the virus from the immune system undermine its ability to replicate.

On the surface of an infected cell, HLA class I molecules show off fragments of viral proteins known as epitopes that alert cytotoxic T cells. Cytotoxic T cells then mobilize an attack on infected cells. How fast an HIV infection progresses depends in part on which versions of HLA class I a person harbors. People with the HLA-B*5703 variant hold virus replication in check longer than people with other HLA alleles. But HIV is slippery. To get around HLA-B*5703, HIV rapidly mutates three residues in its Gag p24 protein.

These mutations erase three of the major epitopes that alert cytotoxic T cells in people with HLA-B*5703, rendering the virus invisible to the immune system. But, as Crawford et al. now show, the virus pays for its invisibility. In culture, the triple mutant viruses replicated 20 times slower than normal.

Despite this handicap, the virus still came out ahead in people with HLA-B*5703. When a virus with only two mutations acquired the third, its abundance in the blood leapt by 10 times. Overall, however, patients with HLA-B*5703 were better off than most—with circulating viral titers about half the average for HIV-infected people.

The team also followed the mutations’ impact after transmission by studying Zambian couples in which one person had infected the other. In recipients that lacked HLA-B*5703, the virus gradually lost the mutations, as the benefit of avoiding killer T cells no longer outweighed the cost of reduced replication. But patients unlucky enough to have the HLA type that the virus was already adapted to avoid, sickened rapidly. These data suggest that vaccines should be designed to produce a T cell response against many epitopes, something that experimental human vaccines so far haven't achieved.