There is currently a huge gap between the amount of experimental data on infectious disease and the availability of good models to interpret it. Here we present a method to create quantitative and predictive models of infectious disease pathways. As example we have developed a comprehensive diagram of the Influenza A life cycle and its interaction with the anti-viral pathways of the host macrophages. Our hope is that our Influenza A Infection model can aid data interpretation, hypothesis generation, drug development and the improved design of animal experiments.
Influenza A Virus (IAV)
Influenza A virus (IAV) belongs to the Orthomyxoviridae family of enveloped viruses. Viral particles measure 80–120 nanometres and can be spherical or filamentous. Its genome consists of eight segments of negative-sense ssRNA (Neumann et al., 2004). The term negative-sense RNA implies that the RNA must first be transcribed to positive-sense RNA before it can be translated into proteins. In Figure 1 we represent a spherical Influenza virion drawn using our mEPN graphical notation system (Freeman et al., 2010). Multimeric surface proteins (HA, blue; NA, pink; M2, green) are represented in mEPN notation as protein complexes (yellow boxes). Genome segments are packaged into the core in eight helical Ribonucleoprotein complexes (RNPs), containing ssRNA and nucleoprotein, NP. Each RNP is associated with 3 polymerase polypeptides (PA1, PA2 and PA). The virus core is coated by the structural matrix protein M1 (purple). The segmented nature of the viral genome allows the exchange of genetic material between different strains (reassortment) during co-infection of cells. Strains are divided based on their distinct surface proteins hemagglutinin (HA) and neuraminidase (NA).
Figure 1 Influenza A spherical virion drawn using the mEPN graphical notation (click to zoom).
Why is it important?
Influenza A virus is a major cause of seasonal epidemics and sporadic pandemics in humans. The virus can also infect other mammals and birds. In humans it causes the Flu disease, an acute respiratory tract infection which makes patients more vulnerable to other infections, such as pneumonia. Seasonal epidemics affect many millions of individuals worldwide resulting in a significant economic burden on health organizations.
In 1918 the H1N1 strain gave rise to a lethal pandemic known as the “Spanish flu”, which caused an estimated 50 million fatalities. The first human influenza virus was isolated after propagation in ferrets in 1933 (Smith et al., 1933) paving the way for developing influenza vaccines. However, antigenic changes and genetic reassortment triggered the 1957, 1968 and 2009 pandemic outbreaks contributing to the evolution of the current swine-derived H1N1 (PDM 2009 H1N1) (Smith et al., 2009).
Current knowledge has been gathered mostly from animal research using adapted laboratory strains, as H1N1 A/Puerto Rico/8/34 (PR8) and A/WSN/33 (WSN), which were derived by serial passage through model animals.
Animal models have been widely used for the study of viral and host factors contributing to influenza virus infection and for the preclinical evaluation of vaccines and antiviral drugs. Ideally the animal model should closely mimic humans in terms of clinical signs and virus growth, but each species has its own limitations. There are, as example, several important differences between the manifestations of influenza in mice and humans (Belser et al. 2010).
There is currently a need for new approaches to overcome the limitations of animal models and gain new insights that can prevent influenza virus outbreaks.
Modelling the Influenza A Virus Life Cycle
We have constructed a large and detailed model of the molecular interactions involved in the Influenza A virus replication cycle (Figure 2). As sources we used Reactome and the Influenza A model from Matsuoka et al. (2013), which in itself contains information gleaned from over 500 papers. We also have frequently revised the primary data during the construction of these models.
Figure 2 Model of the Influenza A Life Cycle drawn using the mEPN notation (click to zoom).
We layed-out the diagram with a modular structure (Figure 3) that facilitates reading and allows for network expansion as new data become available.
Figure 3 Modules illustrating the different phases of IAV replication. Virus present in the extra-cellular space (lilac) enters into the cytoplasm (yellow) and replicates within the nucleus (green). New virions (Virus Output) budding from the cellular membrane (orange) are released at the end of the life cycle.
The diagram was constructed to support computational modeling using SPN Petri net algorithm (Ruths et al. 2008), which simulates system dynamics using tokens accumulation as a measure of component’s activity. To perform the parameterisation we consulted extensive literature about the synthesis of viral components during the IAV life cycle to closely reproduce experimentally observed kinetics. When our virtual cell is challenged by a limited amount of virus (MOI 10 for 2 time blocks, 10 virions are represented by 10 input tokens, 1 time block = 3 min) it produces an amount of viral progeny (~1E4 virions/cell Virus Output, Figure 4) comparable to that obtained by in vitro infection (Sidorenko and Reichl 2004). The relative amount of each viral messenger RNA and its accumulation kinetic were modeled after Shapiro et al. (1987) and Hatada et al. (1989) considering the input of newly produced viral ribonucleoproteins (vRNP) and the inihbitory feeback of the viral M1 protein, which limits transcription. To simulate early and late phases of viral protein production introduced a delay so that packaging and structural proteins (HA,NA, M1 and NS2) accumulate later in the life cycle. (Figure 4). The dynamic accumulation of tokens in the mEPN model can be visualized using BioLayout Express3D to control the animation of the simulation (Video 1).
Figure 4 The accumulation of viral components (mRNA, proteins) observed running IAV life cycle simulations closely reproduces experimental kinetics. Simulation: 500 runs, 100 time blocks, standard normal distribution, 20 time blocks = 1 hour.
Video 1 Animation of the Signaling Petri Net simulation algorithm for the IAV life cycle model. Simulation performed using BioLayout Express3D with the following parameters: 500 runs, 100 time blocks, standard normal distribution. The inflation of the nodes represents the accumulation of the viral components during IAV replication.
Modelling the immune cells response to Influenza A Infection
To generate an integrated model of the Influenza A Infection we combined the IAV life cycle model described above with a map of the anti-viral pathways of macrophages (Raza et al., 2010). We focused on the effector pathways designed to neutralise viruses with particular attention to the Inferferon-β signaling, which plays an important role in innate immunity.
Airway macrophages and inflammatory monocytes play a pivotal role in mounting effective innate and adaptive responses. These immune cells play a critical role in controlling the Influenza A Virus (IAV) infection, removing infectious virions via phagocytosis and secreting inflammatory cytokines.
Several studies have reported that IAV infection is ‘abortive’ in macrophages, which can be infected but don’t produce virions and die within 36 hours (Bender et al.1998, Ioannidis et al. 2012). Abortive infection can contribute to host defece acting a dead-end for low pathogenic strains.
The infection outcome appears to be strain dependent. Some highly pathogenic Influenza strains can replicate productively in macrophages. As example, H5N1 virus replicates in Alveolar Macrophages (AMs) better than H1N1. Both strains can replicate Peripheral Blood Monocyte-Derived Macrophages (PBDMs) (Yu et al. 2011).
Investigating the factors that determine a productive versus abortive infection in macrophages is therefore of paramount importance to face the challenges highly pathogenic influenza strains, as well as many other viruses that infect these cells.
Influenza A Infection Model
We focused on the classical effector pathways that play a role in the neutralization of viruses (Figure 5). We also included newly decribed host factors such as IFITM3 and MX1 that have been shown to limit influenza infection (reviewed by Iwasaki and Pillai 2014).
Figure 5 Modules illustrating the cellular pathways involved in anti-viral response. TLR, Toll-Like Receptors; IFN, Interferon; 25OAS, 2′,5′-oligoadenylate synthetase; PKR, protein kinase R.
The integrated network model comprises of 1,777 nodes and 1,993 edges, which includes 205 nodes representing viral and host proteins and 157 protein complexes (Figure 6).
Figure 6 Model representing the Influenza A Infection of a Macrophage cell (click to zoom).
To evaluate the predictivity of our model we performed virtual Influenza A infection simulations in inflammatory monocytes and Interferon-β primed macrophages. As an example, in human alveolar macrophages infection with the H1N1 strain is abortive, but inflammatory monocytes subsequently recruited to the lung support active viral replication (reviewed by Short et al. 2012) (Figure 7).
Figure 7 Influenza A Virus Infection (H1N1 strain) is productive in epithelial cells and newly recruited inflammatory monocytes. In contrast, Interferon-β primed macrophages do not support viral replication and the infection is abortive.
The simulations were performed for an average 500 runs assuming a standard normal distribution for the token flow and 100 time blocks (roughly equivalent to 5 hours). We compared the Influenza infection in macrophages described by our integrated diagram to the Influenza Life-cycle map that we modelled using data from epithelial cells (Shapiro et al. 1987; Hatada et al. 1989). Endogenous Interferon-β is produced by the macrophage upon TLRs stimulation, however to simulate an Interferon (IFN)-primed macrophage we introduced 1000 input tokens on the IFNB1 protein. As outcome measures we estimated the Virus Output and tokens accumulation on the Apoptosis pathway. Our model predicts that interferon-primed alveolar macrophages are more effective against IAV infection resulting in lower Virus Output and increased Apoptosis in comparison to epithelial cells (yellow) and newly recruited inflammatory monocytes (blue) (Figure 8). Macrophages that have been exposed to Interferon-β produce higher amounts of the antiviral factor IFITM3, which limits the virus entry, and more MX1 that blocks viral replication. Thus, Interferon-β stimulus appears essential for the effective block of Influenza infection as shown in Video 2.
Video 2 Animation of two parallel simulations. Top diagram shows the simulation of a productive IAV infection in an inflammatory monocyte; Bottom diagram shows an abortive IAV infection in Inferferon-β primed macrophage. Simulation performed using BioLayout Express3D with the following parameters: 500 runs, 100 time blocks, standard normal distribution. The inflation of the nodes represents the accumulation of the viral components.
Considering the importance of Interferon-β signalling we sought to test the virtual infection of a host missing the Interferon-responsive factor IFITM3. In our model IFITM3 protein produced upon Interferon stimulation blocks the Virus-Endosome fusion thus reducing the release of viral RNPs into the cytoplasm (Desai et al. 2014). To simulate an Ifitm3 knock-out we removed the Ifitm3 gene from our diagram, thus releasing the restrain on IAV infection. Our model corroborates the notion that influenza infection is productive in Ifitm3 null host, as the infection of IFN-primed macrophages produces an output of ~4000 virions/cell (Figure 8, white line showing IFITM3 knock-out Virus Output). This observation is consistent with the experimental evidence that mice lacking Ifitm3 display fulminant viral pneumonia in response to influenza (Everitt et al. 2012). Similar but less severe outcome was obtained from the knock-out of the Interferon-responsive anti-viral gene Mx1 (Verhelst et al. 2012).
Figure 8 Top panels: comparison of the Virus Output and Pro-apoptotic signalling obtained simulating IAV infection of epithelial cells (yellow), inflammatory monocytes (blue) and Inferferon-β primed macrophages (magenta). Bottom panels: Virus Output from IAV Infection of wild-type Inferferon-β primed macrophages(magenta) in comparison to the output from Ifitm3- (white) and Mx1- null (green) Inferferon-β primed macrophages. Simulation: 500 runs, 100 time-blocks, normal distribution, 20 time blocks = ~1 hour.
- The Influenza A Infection Model distils a wealth of literature into an easy-to-browse diagram, which provides a system level picture of the known interactions between influenza and macrophages.
- This model is an excellent tool to optimize experimental design and generate new hypotheses.
- Scientists can run in silico experiments to avoid redundant or unnecessary wet-lab experiments and reduce the number of animals used to a minimum
Phase 1 of the NC3Rs-funded Crack It Challenge brought together an international team possessing extensive expertise in influenza virus and macrophage biology, bioinformatics, data analyses and visualisation, and graphical and computational pathway modelling.
Prof Tom Freeman
Professor Tom Freeman holds the Chair of Systems Immunology and has a long history of working in the field of functional genomics and systems biology. His work focuses on exploring immune pathways that regulate the molecular interplay between host and pathogen, particularly with respect to the macrophage. He has been active in developing a graphical pathway notation scheme (mEPN) to model biological interactions within or between cells. His team has developed BioLayout Express3D in order to support the integration of pathway modelling and ‘omics data analysis. He is actively involved in a number of projects analysing immune expression signatures, promoter-based analyses and also runs a wet lab to follow up his computational analyses.
Prof Duccio Cavalieri
Professor Duccio Cavalieri obtained his PhD in Genetics from the University of Pavia in 1998 he moved to Harvard. His work on pathway analysis in 2002 was seminal to the application of bioinformatics to the interpretation of large “omics” datasets. In 2004 he became group leader at Florence University, studying the boundaries between commensalism and pathogenicity. Since April 2012 he has been head of the computational biology department at FEM and has taken up the challenge of applying next generation sequencing and pathway analysis to study host-microbe interaction at the systems level. The common theme throughout his work is development and application.
Dr David Lynn
Dr David Lynn
Dr David Lynn has a track record in applying computational and systems biology approaches to investigate the host response to infectious disease. Following a PhD in computational immunology at University College Dublin, he moved to Vancouver where he was the lead computational biologist on a Grand Challenges in Global Health Initiative project to investigate the host response to several key pathogens at a systems level. He continues to lead the development of InnateDB.com, a systems biology platform for innate immunity networks. In 2009 he led the Computational Biology at Teagasc during which time his group has developed an integrative biology approach to investigate the immune response to important bovine pathogens. In 2014 DL was appointed as EMBL Australia Group Leader in Biomedical Informatics at the South Australian Health and Medical Research Institute (SAHMRI).
Prof Paul Digard
Professor Paul Digard holds the Chair of Virology at Edinburgh University. He is a molecular virologist with over 25 years experience of a diverse range of pathogens, but primarily focussed on influenza virus. He did his postdoctoral training at Harvard Medical School before returning to the Dept. of Pathology in Cambridge where he undertook his postgraduate training. He moved to the Roslin Institute in 2012 where his research is focused on the molecular and cellular biology of influenza A virus replication. Longstanding interests concern the control mechanisms of viral nucleic synthesis, antivirals, the intracellular trafficking of viral components and the mechanism of virus assembly. More recently, his laboratory has changed the text-book picture of the influenza A virus coding strategy by defining a suite of previously unrecognised viral accessory proteins with roles in virus replication and pathogenesis.
Prof David Hume
Professor David Hume FRSE, FSB, FMedSci moved to The Roslin Institute from the Institute for Molecular Bioscience at the University of Queensland in 2007. He has an international reputation in the field of macrophage biology, and in genome sciences, with >300 peer reviewed papers. Besides primary research papers, he has published many reviews on different aspects of macrophage biology and genome science, including two that have won the Dolph Adams prize for most cited review in the Journal of Leukocyte Biology. His work was recognized in 2010 by award of a Royal Society Wolfson Research Merit Award, in 2011 with the Bonazinga Award for lifetime achievement by the US Society for Leukocyte Biology, and in 2012 the OMICS Award from RIKEN, Japan.
Derek Wright has been working professionally in software engineering since 2000, in commercial, university and consultancy roles. Derek also has a background in life sciences, with B.Sc. Hons in Botany from University of Glasgow and M.Phil. in Botany from University College London. He is a member of the International Society for Computational Biology. Derek worked as Lead Developer for BioLayout Express3D from 2012 to 2014. Derek has expertise in pathway computational biology, having developed a visualisation system for the BioPAX pathway exchange standard, with mappings to the mEPN graphical notation and a web service client to the Pathway Commons resource. He designed and developed the websites for Virtually Immune and BioLayout Express3D and is webmaster of The Macrophage Community Website. Derek gave a talk on the group’s work on Petri Net pathway simulations at the Combine workshop, Paris (2013) and presented work on pathway visualisation at the Visualizing Biological Data conferences, Boston (2013) and Heidelberg (2014). Derek demonstrated Virtually Immune to the public at The Royal Highland Show 2014. Derek’s current position is Bioinformatician in Professor Andrew Biankin’s group at Wolfson Wohl Cancer Research Centre, University of Glasgow, researching precision medicine for pancreatic cancer.
Dr Alessandra Livigni
Dr Alessandra Livigni
Dr Alessandra Livigni is a stem cell and developmental biologist (MSc Medical Biotechnology (Hons); PhD in Molecular Pathology) with fourteen years of experience. Alessandra’s first post-doc, examined the gene regulatory networks (GRN) controlling tissue patterning and homeostasis, focusing on the transcription factor Oct4 , an essential regulator of stem cells self-renewal and pluripotency. Her work uncovered a novel role for Oct4 as regulator of inter-cellular adhesion and this has important implications for embryonic stem cells differentiation. Alessandra’s current interests are in employing wet-lab and computational approaches to study how signal transduction pathways regulate cell plasticity and homeostasis. In her current post, funded by the NC3Rs Crack it Challenge, Alessandra worked on a biologist-friendly approach to modelling and designed the virtual infection model for Influenza A Virus to aid studying the signalling pathways involved in immune response.
Dr Kenneth Baillie
Dr Ken Baillie is an intensive care doctor and research scientist exploring why some people are susceptible to infections, such as flu. In particular his research interests are in the genetics of host susceptibility to severe infection. He led the GenISIS (Genetics of Influenza Susceptibility in Scotland) study and the host genetics component of the MOSAIC (Mechanisms of Severe Influenza Consortium) study. In his role as working group chair for genomics, pathogenesis and pharmacology for the International Severe Acute Respiratory Infection Consortium (ISARIC), he led the development of an integrated biological sampling protocol for use in outbreaks, which is supported by the World Health Organization and has been adopted in many countries throughout the world.