Reservoirs of Disease

I am interested in what makes some diseases chronic. The disease agent, which may be a virus, bacterium, parasite, or in some cases a cancer cell, is never fully cleared by the immune system or therapy. Instead it persists at low levels as an intra-patient reservoir which has the ability to re-seed widespread disease. In HIV, the reservoir is of critical importance since it persists in infected individuals despite the presence of suppressive antiretroviral therapy, and has so far been an insurmountable barrier to achieve a cure. I want to understand how reservoirs are created, and devise therapeutic strategies that specifically target reservoirs in human disease.

HIV

Current antiretroviral therapy has been successful in reducing HIV to undetectable levels in most patients but not in eradicating the infection, because low level viremia persists for life. Part of the reservoir may have its roots in ongoing viral replication. Why do drugs suppress replication to undetectable levels, but fail to suppress replication completely?

Infection can occur by the classical “cell-free” mode of infection and by cell-to-cell HIV spread, where the latter relies on the infected cell, not the virus, to do the work of finding a new cell to infect. In cell-to-cell spread, the number of virus particles failing to reach the target cell is minimized and so the virus dose delivered per infected cell is potentially large.

I investigated whether cell-to-cell HIV spread can result in HIV transmission which is insensitive to antiretroviral drugs. I constructed a probabilistic model of infection under drug which predicted that if many viruses are transmitted to one cell, the probability that at least one virus escapes the drug and succeeds in infecting the cell is dramatically increased. This is intuitively illustrated in the figure below:

Multiple infections per cell decrease sensitivity to drug. Red circles - infected cells, arrows – transmissions, hexagons or hexagons surrounded by circles–viruses, broken circles –degraded viruses, X – viruses blocked by drug, wavelets –successful infection.

To test the model, I infected cells with cell-free HIV at low and high multiplicity of infection per cell. At high multiplicity, infection was insensitive to therapy at a range of drug concentrations. This shows that a large viral dose per infected target cell is sufficient to confer insensitivity to drugs. I then used antiretroviral drugs to inhibit HIV infection by low multiplicity cell-free virus (which is the physiologically relevant level of cell-free virus) or by cell-to-cell spread. While a similar number of cells were infected by each mode in the absence of drugs, infections originating from cell-free virus decreased strongly in the presence of antiretroviral drugs while infections involving cell-to-cell spread were much less sensitive. The reduction in sensitivity resulted in multiple rounds of infection with cell-to-cell spread in the presence of drugs. If cell-to-cell spread acts in a similar way in vivo, loss of sensitivity to drugs should lead to constant or intermittent low levels of ongoing replication in the face of HIV inhibitors, creating a virus reservoir.

Cancer

My interest in reservoirs started in cancer, where ‘fractional killing’ with anti-cancer drugs is often observed. The drugs kill of most of the targeted cell population, but a small fraction survives to reseed the disease. I wanted to know whether such heterogeneity can be explained by fluctuations in protein states in individual cells. To study this, I constructed a library of proteins labeled with YFP at their native chromosomal locations and monitored protein dynamics in individual living cells using time-lapse microscopy and automated image analysis. I observed that cells higher than average in the level of a given protein eventually became lower and vice versa. This is illustrated below for two diffrent proteins, the ubiquitin specific protease USP7, and the chromatin remodelling factor HMGA2:

Dynamics of cells expressing YFP-tagged endogenous USP7 and HMGA2 across two cell generations. Each line represents the ranked protein level of one cell, normalized to the population mean. Lines are colour-coded for rank at the start of the first cell cycle, with red being high relative to the mean and blue low.

The memory time for a particular protein state differed between proteins and was on the scale of several days, with proteins in the same network having correlated fluctuations. Subsequently, we and others have demonstrated that transient protein differences between cells explain cell fate decisions in the face of cytotoxic stress and during differentiation. Such transient fluctuations may form a cancer reservoir in the presence of therapy.