Reply to “Early Stochastic Dynamics in Human Cytomegalovirus Infection”

REPLY

We appreciate the comments by Wodarz and Rodriguez-Brenes (1) and their interest in our study, in which we modeled natural acquisition of human cytomegalovirus (CMV) infection (2). We agree that understanding the dynamics of the earliest stages of viral acquisition and the factors that determine the establishment of infection are of fundamental importance. Therefore, the authors' in vitro experiments exploring the dynamics of early adenovirus (AdV) infection are noteworthy (3).

In our study of human CMV infection in vivo, we show that a low basic-reproduction number (R₀) and a small number of initially infected cells result in infections that frequently are rapidly extinguished (2). These findings might be entirely compatible with those of Rodriguez-Brenes et al., who used low-multiplicity AdV infection in vitro. They found that the outcome of AdV infection depends on a race between viral replication and the cellular type I interferon response and that half of infections were extinguished while the rest were able to persist and spread rapidly (3). In our study, the effects of local interferon production are implicitly captured by the parameters of our model, specifically through the rates of death of infected cells with or without replicating virus. As we mention in our paper (2), we think that early innate antiviral responses, such as those of type I interferons, are likely to play an important role in whether replication following a given viral exposure results in limited or persistent spread (4). However, our analysis also indicates that inherently weak viral spread is a critical factor during the initial CMV-host interaction in vivo.

It should be noted that there are likely important differences between the early-replication dynamics of AdV and CMV, whether in vivo or in vitro. In contrast to Wodarz and Rodriguez-Brenes, we did not find a substantially higher growth rate at higher viral loads during natural human CMV infection (5). Rather, the protracted exponential growth that we observed during early primary CMV infection predicts an infection system with a low R₀ (2). In other words, our evidence for a low R₀ contributing to stochastic transient infections is derived from the dynamics that we observed during primary infection with high viral loads. As such, we think that the ability of CMV to successfully establish infection despite its remarkably slow viral expansion likely depends more on the numerous strategies that it has evolved to evade the host immune response (6–8) than on a capacity to outrace it.

The early events that determine the outcome of natural infection by low-multiplicity-number viral exposures are almost certainly more complex than mathematical modeling of data from human cohorts or in vitro systems can recapitulate and involve numerous stochastic as well as virus- and host-specific factors. However, continuing to define these elements for a specific virus will be invaluable for the development of new vaccines or viral-vector-based therapies.