A mathematical framework for estimating pathogen transmission fitness and inoculum size using data from a competitive mixtures animal model

Cross-scale dynamics and the evolutionary emergence of infectious diseases

When emerging pathogens encounter new host species for which they are poorly adapted, they must evolve to escape extinction. Pathogens experience selection on traits at multiple scales, including replication rates within host individuals and transmissibility between hosts. We analyze a stochastic model linking pathogen growth and competition within individuals to transmission between individuals. Our analysis reveals a new factor, the cross-scale reproductive number of a mutant virion, that quantifies how quickly mutant strains increase in frequency when they initially appear in the infected host population. This cross-scale reproductive number combines with viral mutation rates, single-strain reproductive numbers, and transmission bottleneck width to determine the likelihood of evolutionary emergence, and whether evolution occurs swiftly or gradually within chains of transmission. We find that wider transmission bottlenecks facilitate emergence of pathogens with short-term infections, but hinder emergence of pathogens exhibiting cross-scale selective conflict and long-term infections. Our results provide a framework to advance the integration of laboratory, clinical, and field data in the context of evolutionary theory, laying the foundation for a new generation of evidence-based risk assessment of emergence threats.

Modeling Competitive Mixtures With the Lotka-Volterra Framework for More Complex Fitness Assessment Between Strains

With increasing resolution of microbial diversity at the genomic level, experimental and modeling frameworks that translate such diversity into phenotypes are highly needed. This is particularly important when comparing drug-resistant with drug-sensitive pathogen strains, when anticipating epidemiological implications of microbial diversity, and when designing control measures. Classical approaches quantify differences between microbial strains using the exponential growth model, and typically report a selection coefficient for the relative fitness differential between two strains. The apparent simplicity of such approaches comes with the costs of limiting the range of biological scenarios that can be captured, and biases strain fitness estimates to polarized extremes of competitive exclusion. Here, we propose a mathematical and statistical framework based on the Lotka-Volterra model, that can capture frequency-dependent competition between microbial strains within-host and upon transmission. As a proof-of-concept, the model is applied to a previously-published dataset from in-vivo competitive mixture experiments with influenza strains in ferrets (McCaw et al., 2011). We show that for the same data, our model predicts a scenario of coexistence between strains, and supports a higher bottleneck size in the range of 35–145 virions transmitted from donor to recipient host. Thanks to its simplicity and generality, such framework could be applied to other ecological scenarios of microbial competition, enabling a more complex and nuanced view of possible outcomes between two strains, beyond competitive exclusion.

Population-level mathematical modeling of antimicrobial resistance: a systematic review

BACKGROUND

Mathematical transmission models are increasingly used to guide public health interventions for infectious diseases, particularly in the context of emerging pathogens; however, the contribution of modeling to the growing issue of antimicrobial resistance (AMR) remains unclear. Here, we systematically evaluate publications on population-level transmission models of AMR over a recent period (2006-2016) to gauge the state of research and identify gaps warranting further work.

METHODS

We performed a systematic literature search of relevant databases to identify transmission studies of AMR in viral, bacterial, and parasitic disease systems. We analyzed the temporal, geographic, and subject matter trends, described the predominant medical and behavioral interventions studied, and identified central findings relating to key pathogens.

RESULTS

We identified 273 modeling studies; the majority of which (> 70%) focused on 5 infectious diseases (human immunodeficiency virus (HIV), influenza virus, Plasmodium falciparum (malaria), Mycobacterium tuberculosis (TB), and methicillin-resistant Staphylococcus aureus (MRSA)). AMR studies of influenza and nosocomial pathogens were mainly set in industrialized nations, while HIV, TB, and malaria studies were heavily skewed towards developing countries. The majority of articles focused on AMR exclusively in humans (89%), either in community (58%) or healthcare (27%) settings. Model systems were largely compartmental (76%) and deterministic (66%). Only 43% of models were calibrated against epidemiological data, and few were validated against out-of-sample datasets (14%). The interventions considered were primarily the impact of different drug regimens, hygiene and infection control measures, screening, and diagnostics, while few studies addressed de novo resistance, vaccination strategies, economic, or behavioral changes to reduce antibiotic use in humans and animals.

CONCLUSIONS

The AMR modeling literature concentrates on disease systems where resistance has been long-established, while few studies pro-actively address recent rise in resistance in new pathogens or explore upstream strategies to reduce overall antibiotic consumption. Notable gaps include research on emerging resistance in Enterobacteriaceae and Neisseria gonorrhoeae; AMR transmission at the animal-human interface, particularly in agricultural and veterinary settings; transmission between hospitals and the community; the role of environmental factors in AMR transmission; and the potential of vaccines to combat AMR.

Characterization of Influenza B Virus Variants with Reduced Neuraminidase Inhibitor Susceptibility

Treatment options for influenza B virus infections are limited to neuraminidase inhibitors (NAIs), which block the neuraminidase (NA) glycoprotein on the virion surface. The development of NAI resistance would therefore result in a loss of antiviral treatment options for influenza B virus infections. This study characterized two contemporary influenza B viruses with known resistance-conferring NA amino acid substitutions, D197N and H273Y, detected during routine surveillance. The D197N and H273Y variants were characterized in vitro by assessing NA enzyme activity and affinity, as well as replication in cell culture compared to those of NAI-sensitive wild-type viruses. In vivo studies were also performed in ferrets to assess the replication and transmissibility of each variant. Mathematical models were used to analyze within-host and between-host fitness of variants relative to wild-type viruses. The data revealed that the H273Y variant had NA enzyme function similar to that of its wild type but had slightly reduced replication and transmission efficiency in vivo The D197N variant had impaired NA enzyme function, but there was no evidence of reduction in replication or transmission efficiency in ferrets. Our data suggest that the influenza B virus variant with the H273Y NA substitution had a more notable reduction in fitness compared to wild-type viruses than the influenza B variant with the D197N NA substitution. Although a D197N variant is yet to become widespread, it is the most commonly detected NAI-resistant influenza B virus in surveillance studies. Our results highlight the need to carefully monitor circulating viruses for the spread of influenza B viruses with the D197N NA substitution.

Effects of Transmission Bottlenecks on the Diversity of Influenza A Virus

We investigate the fate of de novo mutations that occur during the in-host replication of a pathogenic virus, predicting the probability that such mutations are passed on during disease transmission to a new host. Using influenza A virus as a model organism, we develop a life-history model of the within-host dynamics of the infection, deriving a multitype branching process with a coupled deterministic model to capture the population of available target cells. We quantify the fate of neutral mutations and mutations affecting five life-history traits: clearance, attachment, budding, cell death, and eclipse phase timing. Despite the severity of disease transmission bottlenecks, our results suggest that in a single transmission event, several mutations that appeared de novo in the donor are likely to be transmitted to the recipient. Even in the absence of a selective advantage for these mutations, the sustained growth phase inherent in each disease transmission cycle generates genetic diversity that is not eliminated during the transmission bottleneck.

Transmission Bottleneck Size Estimation from Pathogen Deep-Sequencing Data, with an Application to Human Influenza A Virus

The bottleneck governing infectious disease transmission describes the size of the pathogen population transferred from the donor to the recipient host. Accurate quantification of the bottleneck size is particularly important for rapidly evolving pathogens such as influenza virus, as narrow bottlenecks reduce the amount of transferred viral genetic diversity and, thus, may decrease the rate of viral adaptation. Previous studies have estimated bottleneck sizes governing viral transmission by using statistical analyses of variants identified in pathogen sequencing data. These analyses, however, did not account for variant calling thresholds and stochastic viral replication dynamics within recipient hosts. Because these factors can skew bottleneck size estimates, we introduce a new method for inferring bottleneck sizes that accounts for these factors. Through the use of a simulated data set, we first show that our method, based on beta-binomial sampling, accurately recovers transmission bottleneck sizes, whereas other methods fail to do so. We then apply our method to a data set of influenza A virus (IAV) infections for which viral deep-sequencing data from transmission pairs are available. We find that the IAV transmission bottleneck size estimates in this study are highly variable across transmission pairs, while the mean bottleneck size of 196 virions is consistent with a previous estimate for this data set. Furthermore, regression analysis shows a positive association between estimated bottleneck size and donor infection severity, as measured by temperature. These results support findings from experimental transmission studies showing that bottleneck sizes across transmission events can be variable and influenced in part by epidemiological factors.IMPORTANCE The transmission bottleneck size describes the size of the pathogen population transferred from the donor to the recipient host and may affect the rate of pathogen adaptation within host populations. Recent advances in sequencing technology have enabled bottleneck size estimation from pathogen genetic data, although there is not yet a consistency in the statistical methods used. Here, we introduce a new approach to infer the bottleneck size that accounts for variant identification protocols and noise during pathogen replication. We show that failing to account for these factors leads to an underestimation of bottleneck sizes. We apply this method to an existing data set of human influenza virus infections, showing that transmission is governed by a loose, but highly variable, transmission bottleneck whose size is positively associated with the severity of infection of the donor. Beyond advancing our understanding of influenza virus transmission, we hope that this work will provide a standardized statistical approach for bottleneck size estimation for viral pathogens.

Contact transmission of influenza virus between ferrets imposes a looser bottleneck than respiratory droplet transmission allowing propagation of antiviral resistance

Influenza viruses cause annual seasonal epidemics and occasional pandemics. It is important to elucidate the stringency of bottlenecks during transmission to shed light on mechanisms that underlie the evolution and propagation of antigenic drift, host range switching or drug resistance. The virus spreads between people by different routes, including through the air in droplets and aerosols, and by direct contact. By housing ferrets under different conditions, it is possible to mimic various routes of transmission. Here, we inoculated donor animals with a mixture of two viruses whose genomes differed by one or two reverse engineered synonymous mutations, and measured the transmission of the mixture to exposed sentinel animals. Transmission through the air imposed a tight bottleneck since most recipient animals became infected by only one virus. In contrast, a direct contact transmission chain propagated a mixture of viruses suggesting the dose transferred by this route was higher. From animals with a mixed infection of viruses that were resistant and sensitive to the antiviral drug oseltamivir, resistance was propagated through contact transmission but not by air. These data imply that transmission events with a looser bottleneck can propagate minority variants and may be an important route for influenza evolution.

Connecting within-host dynamics to the rate of viral molecular evolution

Viruses evolve rapidly, providing a unique system for understanding the processes that influence rates of molecular evolution. Neutral theory posits that the evolutionary rate increases linearly with the mutation rate. The occurrence of deleterious mutations causes this relationship to break down at high mutation rates. Previous studies have identified this as an important phenomenon, particularly for RNA viruses which can mutate at rates near the extinction threshold. We propose that in addition to mutation dynamics, viral within-host dynamics can also affect the between-host evolutionary rate. We present an analytical model that predicts the neutral evolution rate for viruses as a function of both within-host parameters and deleterious mutations. To examine the effect of more detailed aspects of the virus life cycle, we also present a computational model that simulates acute virus evolution using target cell-limited dynamics. Using influenza A virus as a case study, we find that our simulation model can predict empirical rates of evolution better than a model lacking within-host details. The analytical model does not perform as well as the simulation model but shows how the within-host basic reproductive number influences evolutionary rates. These findings lend support to the idea that the mutation rate alone is not sufficient to predict the evolutionary rate in viruses, instead calling for improved models that account for viral within-host dynamics.

Estimating the fitness advantage conferred by permissive neuraminidase mutations in recent oseltamivir-resistant A(H1N1)pdm09 influenza viruses

Oseltamivir is relied upon worldwide as the drug of choice for the treatment of human influenza infection. Surveillance for oseltamivir resistance is routinely performed to ensure the ongoing efficacy of oseltamivir against circulating viruses. Since the emergence of the pandemic 2009 A(H1N1) influenza virus (A(H1N1)pdm09), the proportion of A(H1N1)pdm09 viruses that are oseltamivir resistant (OR) has generally been low. However, a cluster of OR A(H1N1)pdm09 viruses, encoding the neuraminidase (NA) H275Y oseltamivir resistance mutation, was detected in Australia in 2011 amongst community patients that had not been treated with oseltamivir. Here we combine a competitive mixtures ferret model of influenza infection with a mathematical model to assess the fitness, both within and between hosts, of recent OR A(H1N1)pdm09 viruses. In conjunction with data from in vitro analyses of NA expression and activity we demonstrate that contemporary A(H1N1)pdm09 viruses are now more capable of acquiring H275Y without compromising their fitness, than earlier A(H1N1)pdm09 viruses circulating in 2009. Furthermore, using reverse engineered viruses we demonstrate that a pair of permissive secondary NA mutations, V241I and N369K, confers robust fitness on recent H275Y A(H1N1)pdm09 viruses, which correlated with enhanced surface expression and enzymatic activity of the A(H1N1)pdm09 NA protein. These permissive mutations first emerged in 2010 and are now present in almost all circulating A(H1N1)pdm09 viruses. Our findings suggest that recent A(H1N1)pdm09 viruses are now more permissive to the acquisition of H275Y than earlier A(H1N1)pdm09 viruses, increasing the risk that OR A(H1N1)pdm09 will emerge and spread worldwide.

The experimental aerosol transmission of influenza virus

The threat of a virulent, highly transmissible pandemic virus has motivated an escalating research effort to identify the transmissible genotypes of animal viruses that cross over into the human population (animal–human transmission) and sustain human–human transmission. In addition to the pursuit of the viral genotype, a greater understanding of the host-virus phenotype of infectiousness, transmissibility and susceptibility will be required. This review examines experimental animal transmission of influenza for insights into human influenza transmission. Transmission is viewed as sequential steps that the virus must pass critical thresholds to achieve transmission and ultimately survival in the human host. In particular, a quantitative understanding in animal models of viral replication efficiency, airway viral load, exhaled viral aerosol load, environmental virus survival and host susceptibility will likely yield important insights. Computational modeling will enhance animal model data, as well as guide the use of pandemic mitigation strategies.

Drivers and consequences of influenza antiviral resistant-strain emergence in a capacity-constrained pandemic response

Antiviral agents remain a key component of most pandemic influenza preparedness plans, but there is considerable uncertainty regarding their optimal use. In particular, concerns exist regarding the likelihood of wide-scale distribution to select for drug-resistant variants. We used a model that considers the influence of logistical constraints on diagnosis and drug delivery to consider achievable 'reach' of alternative antiviral intervention strategies targeted at cases of varying severity, with or without pre-exposure prophylaxis of contacts. To identify key drivers of epidemic mitigation and resistance emergence, we used Latin hypercube sampling to explore plausible ranges of parameters describing characteristics of wild type and resistant viruses, along with intervention efficacy, target coverage and distribution capacity. Within our model framework, 'real world' constraints substantially reduced achievable drug coverage below stated targets as the epidemic progressed. In consequence, predictions of both intervention impact and selection for resistance were more modest than earlier work that did not consider such limitations. Definitive containment of transmission was unlikely but, where observed, achieved through early liberal post-exposure prophylaxis of known contacts of treated cases. Predictors of resistant strain dominance were high intrinsic fitness relative to the wild type virus, and early emergence in the course of the epidemic into a largely susceptible population, even when drug use was restricted to severe case treatment. Our work demonstrates the importance of consideration of 'real world' constraints in scenario analysis modeling, and highlights the utility of models to guide surveillance activities in preparedness and response.

Viral fitness: definitions, measurement, and current insights

Viral fitness is an active area of research, with recent work involving an expanded number of human, non-human vertebrate, invertebrate, plant, and bacterial viruses. Many publications deal with RNA viruses associated with major disease emergence events, such as HIV-1, influenza virus, and Dengue virus. Study topics include drug resistance, immune escape, viral emergence, host jumps, mutation effects, quasispecies diversity, and mathematical models of viral fitness. Important recent trends include increasing use of in vivo systems to assess vertebrate virus fitness, and a broadening of research beyond replicative fitness to also investigate transmission fitness and epidemiologic fitness. This is essential for a more integrated understanding of overall viral fitness, with implications for disease management in the future.

Multidrug resistant 2009 A/H1N1 influenza clinical isolate with a neuraminidase I223R mutation retains its virulence and transmissibility in ferrets

Only two classes of antiviral drugs, neuraminidase inhibitors and adamantanes, are approved for prophylaxis and therapy against influenza virus infections. A major concern is that influenza virus becomes resistant to these antiviral drugs and spreads in the human population. The 2009 pandemic A/H1N1 influenza virus is naturally resistant to adamantanes. Recently a novel neuraminidase I223R mutation was identified in an A/H1N1 virus showing cross-resistance to the neuraminidase inhibitors oseltamivir, zanamivir and peramivir. However, the ability of this virus to cause disease and spread in the human population is unknown. Therefore, this clinical isolate (NL/2631-R223) was compared with a well-characterized reference virus (NL/602). In vitro experiments showed that NL/2631-I223R replicated as well as NL/602 in MDCK cells. In a ferret pathogenesis model, body weight loss was similar in animals inoculated with NL/2631-R223 or NL/602. In addition, pulmonary lesions were similar at day 4 post inoculation. However, at day 7 post inoculation, NL/2631-R223 caused milder pulmonary lesions and degree of alveolitis than NL/602. This indicated that the mutant virus was less pathogenic. Both NL/2631-R223 and a recombinant virus with a single I223R change (recNL/602-I223R), transmitted among ferrets by aerosols, despite observed attenuation of recNL/602-I223R in vitro. In conclusion, the I223R mutated virus isolate has comparable replicative ability and transmissibility, but lower pathogenicity than the reference virus based on these in vivo studies. This implies that the 2009 pandemic influenza A/H1N1 virus subtype with an isoleucine to arginine change at position 223 in the neuraminidase has the potential to spread in the human population. It is important to be vigilant for this mutation in influenza surveillance and to continue efforts to increase the arsenal of antiviral drugs to combat influenza.

Recently, a 2009 pandemic A/H1N1 influenza virus was isolated from an immune compromised patient, with antiviral resistance to the neuraminidase inhibitor class of drugs. This virus had an amino acid change in the viral neuraminidase enzyme; an isoleucine at position 223 was substituted for an arginine (I223R). Patients infected with a pandemic virus that is resistant to all neuraminidase inhibitors, would leave physicians without antiviral treatment options, since these viruses are naturally resistant to the other class of antivirals, the adamantanes. To date, it is unknown if this I223R mutant virus is affected in its ability to cause severe disease and to transmit to other humans. Therefore, we have addressed this question by comparing the I223R mutant virus with a wild type reference virus in a ferret pathogenicity and transmission model. We found that the I223R mutant virus was not severely affected in its pathogenicity, although fewer lung lesions and alveolitis scores were found for the I223R mutant virus. In addition, we demonstrated that this virus transmitted efficiently to naïve ferrets. Consequently, we conclude that this I223R mutant virus has the potential to cause disease and may spread among humans. Therefore, influenza surveillance for this resistance pattern is advised.