Predicting the pathogenesis of influenza from genomic response: a step toward early diagnosis

Infection with influenza virus does not always lead to symptomatic illness, but it is not currently possible to predict who will be severely affected and who will have mild or no symptoms. Gene expression profiling of biofluids might unlock the complex dynamics of response to acute respiratory virus infections such as influenza. A recent article by Alfred Hero and colleagues used transcriptional microarray analyses to follow the response to symptomatic and asymptomatic influenza infection over time, and revealed a role for type I IFN (IFNβ and IFNα) signaling and the NLRP3 inflammasome in determining the outcome in human infections.

Influenza viruses are RNA viruses and are a common cause of respiratory infection, with seasonal type A and B influenza viruses reaching peak prevalence during the winter months. As the 2009 H1N1 pandemic influenza and the sporadic human deaths from H5N1 highly patho genic avian influenza illustrate, the risk of novel patho genic strains emerging from animal reservoirs is ever present. Understanding of the molecular bases of viral pathogenesis, human susceptibility and immune res ponses is crucial for the development of new vaccines, targeted diagnostics and antiviral drugs.
Human exposure to influenza viruses can lead to a wide variety of pathological outcomes. Whereas some people succumb to symptomatic (Sx) infection, others apparently show few symptoms. Severe influenza A virus infections can result in pneumonia and immuno pathology charac ter ized by high oxidative stress, high levels of proinflam matory chemokines and cytokines (hypercytokinemia), and acute respiratory distress syndrome [1,2]. The influenza virus is cleared mainly through the generation of cytotoxic T lymphocytes (CTLs) and neutralizing anti bodies, and future infections are prevented by adaptive immune responses that are capable of targeting antigeni cally similar strains. Neutralizing antibodies, which are generated during natural infection and by currently approved influenza vaccines, primarily target surface glycoproteins, hemagglutinin and neuraminidase and are strongly associated with protection. However, the first response to influenza virus infection is mediated by the innate immune system.
The early host innate immune response is triggered by recognition of the viral RNA by pathogenrecognition receptors (PRRs; such as RigI, TLR3/7 and NLRP3). Responding primarily in antigenpresenting dendritic cells resident in the respiratory tract, RigI and TLR3/7 induce type I IFN (IFNβ and IFNα). Induction of IFNβ/α leads to induction of antiviral, interferonstimulated genes and cytokine/chemokine signals that recruit innate immune effector cells and adaptive immune mechanisms [3,4]. In influenza virus infection, oxidative stress also activates the NLRP3 inflammasome, which is an oligo meric complex of NLRP3 that senses cellular 'danger signals' and possibly viral RNA [5]. NLRP3 contains a caspase activation and recruitment domain that activates caspase 1, which in turn leads to the proteolytic proces sing and release of proinflammatory cytokines IL1β and IL18 [6]. The NLRP3 inflammasome facilitates influenza virus clearance by mediating recruitment of innate immune cells to the lung, and is required for lung repair and survival in mice [5,7].
Microarray analyses of host responses to experimental influenza virus infection in animal models have yielded an unprecedented look into the complex molecular dynamics regulating viral pathogenesis, innate and adaptive immune responses, and inflammation. However, the transcriptional dynamics of host responses to influenza virus infections in humans have been less well studied. To address this question, Huang et al. [8] undertook a novel set of clinical genomics experiments, reported in PLoS Genetics, which we discuss below.

Dynamic response to infection
Huang et al. [8] recruited human volunteers and infected them with the human seasonal influenza strain A/ Abstract Infection with influenza virus does not always lead to symptomatic illness, but it is not currently possible to predict who will be severely affected and who will have mild or no symptoms. Gene expression profiling of biofluids might unlock the complex dynamics of response to acute respiratory virus infections such as influenza. A recent article by Alfred Hero and colleagues used transcriptional microarray analyses to follow the response to symptomatic and asymptomatic influenza infection over time, and revealed a role for type I IFN (IFNβ and IFNα) signaling and the NLRP3 inflammasome in determining the outcome in human infections.
Wyoming/67/2005 (H3N2). The authors mapped the peripheral blood cell transcriptional responses using micro arrays before and during the course of infection (0 to 108 h postinfection). Importantly, all volunteers were immunologically naïve for H3N2specific antibodies and had not been vaccinated for at least 3 years.
Significant differences in transcriptional profiles between Sx and subclinical/asymptomatic (Asx) infection were found. Applying rigorous statistical methods, includ ing unsupervised Bayesian analysis and clinically informed pathway enrichment, Huang et al. mapped the temporal dynamics of clusters of genes and corres ponding cellular pathways. In Sx subjects, influenza virus significantly induced antiviral, inflammatory and immune response gene clusters over the course of an infection [8]. They also detected distinct profiles in Asx patients.

The inflammasome is a key to host response
There is substantial evidence that influenza virus infec tion causes the induction of specific innate immune, inflammatory and oxidative stress pathways in lung tissue and peripheral blood cells in humans and animal models [1,9,10]. Induction of the NLRP3 inflammasome path ways in particular correlate with the onset of symptoms, and are highly upregulated in lung tissue in lethal 1918 influenza virus infection in macaques [1]. Similarly, Huang et al. found that symptomatic influenza virus infection induces a gene signature in peripheral blood cells that is enriched in pathogen recognition receptor, type I IFN, inflammasome and proinflammatory cyto kine response genes [8]. While induction of these path ways was not unexpected, the novelty of this study lies in comparison with Asx subjects.
In contrast to Sx infection, suppression of the inflam masome by limiting oxidative stress may dictate a success ful early response to infection, leading to Asx exposure. Huang et al. present substantial evidence that the Asx group showed 'subclinical' influenza virus infec tion (50% shed virus, 25% showed seroconversion, and all showed similar expression changes in 3,000 genes) [8]. This suggests that the Asx phenotype was evidence of host limitation of influenza virus infection rather than a failed inoculation. Asx subjects showed differential upregulation of antioxidant genes (SOD1 and STK25) and

Limited or no virus infection
Downregulation of NLRP3 SOD1, STK25 antioxidants

Asymptomatic (Asx)
Adaptive immune mechanisms Weak induction of PRR Bortz and García-Sastre Genome Medicine 2011, 3:67 http://genomemedicine.com/content/3/10/67 suppressor of cytokine genes (SOCS2 and SOCS5), and suppression of NLRP3. Taking into account that the peripheral blood cells from which the gene expression profile were taken are at a distance from the site of influenza virus replication (the upper respiratory tract and lungs), the nature of the Asx phenotype is suggestive of an 'immediateearly' decision point upon influenza virus infection of the respiratory tract ( Figure 1). Control of oxidative stress and suppression of the NLRP3 inflammasome may directly limit viral replication, or correlate with induction of a potent antiviral response that targets virusinfected cells before enough viral RNA is produced and recognized by PRR to lead to widespread induction of innate immune pathways such as IFNβ/α.
However, there is a confounding possibility. Asx subjects may have been able to mount a CTLmediated immune response upon infection; it is possible that Asx subjects might only have been immunologically naïve in terms of H3N2 antibody titers. Protein synthesis machi nery was strongly induced in peripheral blood cell RNA of Asx subjects, and there were significant increases in peripheral leukocyte counts [8]. Despite this, influenza specific CTL activity was not measured, so it is unknown whether the Asx phenotype was dependent on the presence of circulating, respiratorytractresident or lungresident Tcell memory targeting conserved cross reactive influenza virus epitopes. Indeed, CTLs raised (in mice) against H3N2 virus are capable of recognizing epi topes in internal proteins (nucleoprotein and polymerase) of a divergent avian H5N1 virus, and contributing to a crossprotective, heterosubtypic response [11].

Genomics leads the way
Besides investigating the nature of the immediateearly repression of virus replication observed in Asx subjects, future genomics studies in human and animal models of influenza virus infections can apply similarly innovative approaches to genomics and clinical/physiological data analysis [8] to yield deeper insight into differences in gene expression signatures in distinct leukocyte subsets in peripheral blood. How do gene expression patterns in peripheral blood cell subsets compare with those in cell populations (epithelial, endothelial, lymphocytic, dendritic cells and macrophages) in the respiratory tract [9]? From a practical view, it is likely that biomarkers that differ entiate early from late symptomatic influenza virus infec tion could be obtained by analyzing peripheral blood [8]. Advancing genomicsbased diagnostic methods may not only contribute to understanding the molecular events leading to symptomatic influenza in human patients, but may also aid the development of pointofcare diagnostics for patients and their close contacts for the early initiation of antiviral treatment.

Competing interests
The authors declare that they have no competing interests.