Exploring virus release as a bottleneck for the spread of influenza A virus infection in vitro and the implications for antiviral therapy with neuraminidase inhibitors

Abstract

Mathematical models (MMs) have been used to study the kinetics of influenza A virus infections under antiviral therapy, and to characterize the efficacy of antivirals such as neuraminidase inhibitors (NAIs). NAIs prevent viral neuraminidase from cleaving sialic acid receptors that bind virus progeny to the surface of infected cells, thereby inhibiting their release, suppressing infection spread. When used to study treatment with NAIs, MMs represent viral release implicitly as part of viral replication. Consequently, NAIs in such MMs do not act specifically and exclusively on virus release. We compared a MM with an explicit representation of viral release (i.e., distinct from virus production) to a simple MM without explicit release, and investigated whether parameter estimation and the estimation of NAI efficacy were affected by the use of a simple MM. Since the release rate of influenza A virus is not well-known, a broad range of release rates were considered. If the virus release rate is greater than ∼0.1 h−1, the simple MM provides accurate estimates of infection parameters, but underestimates NAI efficacy, which could lead to underdosing and the emergence of NAI resistance. In contrast, when release is slower than ∼0.1 h−1, the simple MM accurately estimates NAI efficacy, but it can significantly overestimate the infectious lifespan (i.e., the time a cell remains infectious and producing free virus), and it will significantly underestimate the total virus yield and thus the likelihood of resistance emergence. We discuss the properties of, and a possible lower bound for, the influenza A virus release rate.

A mathematical model describing the localization and spread of influenza A virus infection within the human respiratory tract

“…Figure 11 explores the impact of antiviral therapy with neuraminidase inhibitors (NAIs), administered at various times post-infection, on the time courses for infection with seasonal or avian IAV strains. Since NAIs block the release of newly produced virions, their effect is implemented in the MM, as elsewhere [8,11,[20][21][22]31,45,54], as a reduction in the virus production rate, namely (1−ε NAI )p, from the time treatment is administered, where ε NAI ∈ [0, 1] is the drug efficacy. The efficacy of NAIs was set to ε NAI = 0.98 to study the case of treatment with a high efficacy.…”

Inhibition of sialidase activity as a therapeutic approach

“…A simple mathematical model used to study the inhibition of virus release by sialidase inhibitors was modified to overcome the challenge of including the virus release in the model. 45 To cope with overparameterization resulting from the addition of an explicit release rate to the model, the study considered a range of possible values of the release rate of influenza A virus. The simple model was compared against its variation that included an explicit term for virus release.…”

“…However, the drawback of the proposed strategy is that direct measurement of the virus release rate is difficult. 45 Finally, the rate of viral resistance emergence in influenza patients remains to be determined. 46 …”

Investigating Different Mechanisms of Action in Combination Therapy for Influenza

“…Mathematical models can help in the effort to find optimal combination therapy doses. Within host mathematical models of influenza have previously been used to study many aspects of antiviral treatment including extracting of drug efficacy parameters (Beauchemin et al, 2008 ; Brown et al, 2011 ; Beggs and Dobrovolny, 2015 ; Liao et al, 2017 ), treatment of severe influenza (Dobrovolny et al, 2010 , 2011 ; Deecke and Dobrovolny, 2018 ), emergence of drug resistance (Handel et al, 2007 ; Perelson et al, 2012 ; Hur et al, 2013 ; Canini et al, 2014 ; Dobrovolny and Beauchemin, 2017 ; Deecke and Dobrovolny, 2018 ), and to optimize antiviral treatments (Perelson et al, 2012 ; Heldt et al, 2013 ; Hur et al, 2013 ; Canini et al, 2014 ). While there are some mathematical models that attempt to model infections in patients by including an immune response (Dobrovolny et al, 2013 ; Cao and McCaw, 2015 ; Cao et al, 2015 ; Price et al, 2015 ; Zarnitsyna et al, 2016 ; Yan et al, 2017 ), the lack of appropriate human data for parameterizing and validating these models limits their predictive ability (Dobrovolny et al, 2013 ; Boianelli et al, 2015 ).…”