Fit for a Cure

In his work toward understanding the infection dynamics of HIV, the Laboratory’s Alan Perelson works exclusively with model patients. Literally. He builds computer models of HIV infection to help make sense of puzzling clinical data.

When asked whether he is a virologist or a mathematician, Perelson quips, “Yes.” Pressed for more, he clarifies, “I am a mathematical modeler of viral systems and immune processes.”

Experimental and clinical studies provide the numbers, such as the number of HIV-susceptible cells in a volume of blood or the average number of viruses circulating in the blood. Perelson takes those numbers and devises complex mathematical equations to describe their relationships to one another, then fits his model to the data. If it’s a good fit, he can illuminate portions of the virus-host relationship that are otherwise murky. And during the past year, his team has made some compelling new discoveries.

When a French study reported that a dozen or so patients had quit antiviral drug therapy yet maintained undetectable virus levels—representing a functional cure to an incurable disease—Perelson wanted to know why. If researchers can understand how that happened, then clinicians might be able to help more patients achieve a functional cure. Perelson and postdoctoral researcher Jessica Conway hatched a theory that these patients, who had each been diagnosed and treated very soon after infection, had in so doing given their immune systems a leg up, which kept down the number of viruses in their blood.

By including immune cells in their model, Perelson and Conway determined that an early diagnosis and prompt initiation of treatment can lead to a smaller “latent reservoir”—a hidden population of infected cells. These cells, as long as they aren’t producing more viruses, are invisible to the immune system. But they can become active later and suddenly begin producing viruses. The more cells there are in the latent reservoir, the harder it is for the immune response to contain them once they start producing viruses. Very early treatment for the study patients capped their virus numbers at low levels, keeping their latent reservoirs small, thus allowing their immune systems a chance to get ahead of the infection.

The next question Perelson wanted to answer had to do with the nature of the immune response. Certain types of antibodies, called broad neutralizing antibodies (BnAbs), are effective at preventing the spread of infection from one cell to the next. But because BnAbs only appear years into an HIV infection, their utility is usually handily overwhelmed by the virus. Perelson wanted to find a way to make BnAbs appear early in the infection, when they can do the most good.

To do this, he and postdoctoral researcher Shishi Luo built a model of virus-antibody coevolution. HIV mutates liberally to evade neutralization by antibodies. Similarly, the immune system uses mutation to produce a spectrum of antibodies in hopes that some of them will be strongly matched to the virus. The result is an evolutionary arms race with both sides trying to stay nimble while casting a wide net. The model revealed that antibody production is a zero-sum game, with BnAbs coming at a cost to other types of antibodies and vice versa.

“So, if you’re going to put all your antibody eggs in one basket,” Perelson says, “it had better be the right basket.”

The best antibody basket is indeed BnAbs if—and it’s a big “if”—the BnAbs come along early enough. The way to do that is through viral genetic diversity. More variation early on—say, from an intentionally diverse vaccine preparation—leads to more BnAbs sooner, shifting the infection dynamic in favor of the host.

But Perelson isn’t concentrating solely on the host response. Recently he and postdoctoral researcher Ruian Ke, along with several external collaborators, have been exploring ways of working the other side of the arms race: the latent reservoir, that sleeper cell of sleeper cells. A popular strategy, dramatically called “shock and kill,” attempts to first stimulate latently infected cells into becoming productive and then target them for destruction. Recent clinical results from a latency-reversing agent (LRA) called Vorinostat were inconsistent, with varying extents and durations of activation. Perelson’s task was again to suss out why such variation occurred and also to quantify the impact of Vorinostat. The model that fit the clinical data put LRA-activated cells into a different category than ordinary virus-producing cells and also allowed them to return to latency after a period of time, which makes Vorinostat less an agent of “shock and kill” and more one of “surprise and confuse.”

So Vorinostat turns out to be a lackluster LRA. But when new, better LRAs come along, Perelson and his models will be set to crunch the numbers. Since drug discovery and clinical trials take years, it’s a relief to know that, for model patients at least, a cure may be just a few clicks away.