Systems biology of host-fungus interactions: turning complexity into simplicity.

Modeling interactions between fungi and their hosts at the systems level requires a molecular understanding both of how the host orchestrates immune surveillance and tolerance, and how this activation, in turn, affects fungal adaptation and survival. The transition from the commensal to pathogenic state, and the co-evolution of fungal strains within their hosts, necessitates the molecular dissection of fungal traits responsible for these interactions. There has been a dramatic increase in publically available genome-wide resources addressing fungal pathophysiology and host-fungal immunology. The integration of these existing data and emerging large-scale technologies addressing host-pathogen interactions requires novel tools to connect genome-wide data sets and theoretical approaches with experimental validation so as to identify inherent and emerging properties of host-pathogen relationships and to obtain a holistic view of infectious processes. If successful, a better understanding of the immune response in health and microbial diseases will eventually emerge and pave the way for improved therapies.

Current Opinion in Microbiology 2012, 15:440–446

This review comes from a themed issue on Host–microbe interactions: Fungi

Edited by Mihai G Netea and Gordon D Brown

For a complete overview see the and the

Available online 19th June 2012

1369-5274/$ – see front matter, © 2012 Elsevier Ltd. All rights reserved.

Introduction

In ecology and immunology, tolerance usually refers to host mitigation of the fitness costs of an infection [1]. This is distinct from resistance, whereby the host reduces the microorganism burden. These costs may tip the balance of an immune response towards tolerance of environmental microorganisms, including fungi. Modern pressures on the immune system and the natural composition human microbiome have partially resulted from the expansion of fungi in fermented foods, including opportunistic pathogens colonizing humans. This is particularly important for intestinal tissues, where mucosal immunity faces life-long challenges by beneficial and detrimental microbes [2,3]. These microbes, including pathogenic fungi, possess a molecular arsenal to escape diverse defense mechanisms of immunocompetent hosts. It is thought that the co-evolution of opportunistic pathogens with their healthy host may aid in their ability to exploit host defenses and remain tolerated [4].

The history of host and fungal interactions will strongly influence resistance against and tolerance to microorganisms. Cross-talk mechanisms during host–pathogen interactions will impact the outcome of infections and further influence subsequent pathogen exposure. As a result, genome-wide studies have gained in popularity to investigate global response patterns to infections from both the host and pathogen side. However, biological interpretations of genome-wide studies are limited to only a fraction of the theoretically possible interactions between genes, environmental conditions, and life cycles taking part in a host–pathogen setting (Figure 1). The enormous complexity underlying the host–pathogen interplay when considering the theoretically possible genetic interactions of even a few genes, necessitates the simplification of systems to cellular or pathway levels. A systems biology approach at different levels — genomic, proteomic, and metabolomic — is an emerging strategy to better understand the pathophysiology of infectious processes and their underlying mechanisms during host–pathogen interactions [5,6].

Figure 1

The factorial complexity of host–pathogen interactions. Schematic representation of the number of possible interactions between limited numbers of host–pathogen genes, where interactions are represented as a line between objects. The numbers of potential interactions between genes show a factorial increase based on the number of environmental conditions applicable. The number of putative interactions is calculated by taking the factorial n! of the total number of genes and environmental conditions in the system. This astronomical number for a limited number of genes emphasizes the necessity to simplify host–pathogen interaction studies.

Systems biology is a rapidly evolving integrative approach that connects many disciplines and aims to create a quantitative and predictive understanding of biological processes. Systems biology has evolved by two parallel approaches: ‘top-down’ network inference, reconstruction, and modeling based on functional genomics data, and the ‘bottom-up’ approach of modeling well-defined circuits based on their functional conservation with other systems. Systems biology approaches follow iterative cycles of modeling and data generation, based on a given biological and testable hypothesis [7]. Recent seminal reviews highlight the power of these different approaches in the dissection of mammalian innate immunity [8], the reconstruction of immune signaling, transcriptional networks [13], and host–pathogen interactions [14,5,6].

This review will address recent work aimed at investigating the transition of opportunistic fungi, with a focus on Candida spp., from the commensal to the pathogenic state, emphasizing fungal mechanisms to escape host immune surveillance. We will discuss new approaches in functional genomics that facilitate modeling, and those which are aimed at understanding the fungal response in the host environment. Furthermore, we discuss the advantages of combining different approaches to gain a better understanding of how the cross-talk between fungal pathogens and their hosts shapes the progress and outcome of invasive infections.

Host perspectives

Innate and adaptive immune responses are responsible for recognizing, responding, and adapting to opportunistic microbial pathogens, including fungi [15]. These responses determine whether microbes require the activation of pathogen-specific defense or attack mechanisms [17,18]. Recognition of fungal pathogens by innate immune cells elicits immune responses by engaging multiple cell-bound, soluble, or intracellular receptors, in a stage-specific and cell-specific manner [19]. To date, hundreds of proteins and genes have been implicated in the innate immune response [20]. The transcriptional response to a microbial stimulus is further tailored to both the stimulus and the responsive immune cell [21]. The analysis of the transcriptome of human dendritic cells (DCs) to Aspergillus fumigatus, C. albicans, and S. cerevisiae showed how the expression of immune-relevant genes increases depending on the morphology, life-stage, and incubation period with the fungus [22]. Models of downstream signal transduction networks using gene expression data have been generated based on similarities in expression profiles in related species, the prediction of shared regulatory motifs, and their integration at the pathway level [25]. Recently, a combination of a forward-genomics and reverse-genomics approach enabled the reconstruction of transcriptional and regulatory networks driving the immune response in DCs to a viral infection [26]. The resultant network model investigated how pathogen-sensing pathways achieve specificity and the influence of a single regulator on mediating inflammatory genes and viral responses depending on the timing of the regulator activation. A regulatory network of potential interactions between microRNAs and mRNAs is an additional level of complexity of how pathogens could manipulate host cell responses [27].

The extent to which early transcriptional regulatory events determine the decision-making process in immune cells responding to different pathogenic fungi is still an open question. However, an increasing number of databases are collecting and annotating functional information. For example, the InnateDB, curates the innate immunity interactome [28], and ImmGen collects immunological microarray data (www.immgen.org). The further development of cell-specific bioinformatic tools to analyze the response in macrophages [29] or DCs [30] will allow for the classification of stimuli by their species-specific transcriptional programs governing fungal recognition (Rizzetto et al., unpublished observation).

While the analysis of gene expression is commonly used to study the activation of immune cells, proteomics constitute a complementary approach providing a direct view on protein levels as well as their activities. Proteomics however poses additional challenges, including cost and the technical limitations to make the process quantitative [31]. Moreover, mRNA expression levels are not necessarily correlated with protein production, hampering the comparative analysis of these data sets. A recent study combined a comprehensive quantitative proteome and transcriptome analysis on immature and cytokine-matured human DCs [32]. Although the overall correlation between differential mRNA and protein expression was low, the correlation between components of DC relevant pathways was significantly higher, underscoring that the integration of related data sets at the pathway level can significantly increase the predictive power of multiple -omics analyses. Recently, a global investigation of the macrophage phosphoproteome and its dynamic changes upon TLR activation has been identified [33]. Functional bioinformatic analyses confirmed already known players of the TLR-mediated signaling and identified new transcriptional regulators previously not implicated in TLR-induced gene expression.

Pathogen perspectives

Fungal adaptation to host immune surveillance

Fungal pathogens have developed sophisticated means to evade or persist in the host, despite normal immune surveillance [34]. The use of genome-wide technologies to study global transcriptional changes has revealed the complexity of fungal adaptation to various host niches. Recent studies provide insights into the mechanisms of adaptation during infection, which include: the expression of anti-phagocytic functions and specific nutrient acquisition systems, the remodeling of central carbon metabolism, and the hypoxia response [35,36]. Virulence factor expression is, to a large extent, embedded in the regulation of functions needed for growth in the mammalian hosts. Pioneering early work on the differential gene expression of fungi phagocytosed by immune cells including macrophages, neutrophils, and granulocytes, revealed, among others, a dynamic response to nutrient starvation, oxidative stress, and iron limitation. Attempts by fungi and especially Candida spp. to adapt to the damaging effects of the environment via the activation of genes encoding antioxidant and detoxifying enzymes, and iron uptake proteins were shown [37]. A physiological role for cell surface superoxide dismutases in detoxifying reactive oxygen species (ROS) in innate immune cells and facilitating immune evasion was found [38]. In addition, autophagy and pexophagy mechanisms are important virulence traits of fungi to enable persistence and survival [39,40]. Notably, a global model of iron homeostasis in A. fumigatus has integrated data from Northern blot analysis, microarray expression, transcription factor knock-out mutants, and the occurrence of transcription factor binding motifs in regulatory regions of the genes to predict new transcription factor to target interactions [41].

Fungi may also evade the immune system by changing virulence gene expression at different infection stages upon encountering host-conditions. For example, a novel flow cytometry-based technique showed how changes in fungal gene expression profiles occurring over time influenced patient outcomes with clinical strains of Cryptococcus neoformans [42,43]. Using an in vitro oral candidiasis model, C. albicans mutants defecting in regulators of hyphal formation were attenuated in their ability to invade and damage epithelial cells [44]. The further use of microarray and RNA-seq technology in conjunction with in vitro infection models could be used to further investigate the role of stage-specific virulence gene expression.

Genome dynamics of fungal pathogens

Many fungal clinical isolates display a large degree of genetic and genomic heterogeneity. Segmental or whole-chromosome aneuploidy can be a source of selectable phenotypic variation in fungal species [45], conferring a selective advantage in a host setting [46]. For example, exposure to specific antifungal drugs increases the frequency of adaptive events, promoting drug resistance in independent lineages of C. albicans cells [47]. Additionally, loss of heterozygosity events is elevated in C. albicans in response to oxidative, heat, and antifungal drug stress in vitro [48]. Although rare, even S. cerevisiae may become an opportunistic pathogen under very specific conditions or genetic alterations [49]. Hence, cell population dynamics and evolutionary forces imposed by host stress and other factors may represent the driving force of genomic plasticity in fungal pathogens that enable colonization of various host niches. Strain variability and surface alterations could also explain differences in the host immune response [50], providing new opportunities to model host immune system interactions. Pathogenicity itself could reflect adaptive advantages conferred by the acquisition of virulence traits in different strains, thereby increasing pathogen fitness.

Contrary to S. cerevisiae, C. albicans lacks a complete sexual cycle, impeding efficient genetic analyses and limiting systems biology approaches with this obligatory diploid fungus. Under certain environmental conditions, C. albicans can switch from to a mating-competent state [51]. This transition modulates metabolic preferences, antifungal drug resistance, niche distribution, and host immune cell-specific interactions among many others, and is therefore an important consideration in the investigation of fungal fitness within host niches. Comparative genomics studies have the potential to identify new virulence-associated gene networks [5,52]. The number of sequenced fungal genomes publically available has significantly expanded in recent years [53]. In addition, the Candida Genome Database (CGD) and the Aspergillus database, among others now offer multiple species, facilitating these comparisons. The availability of genomic datasets studying specifically host–fungi interactions have also expanded (Table 1), along with the number of software platforms available for the analysis and integration of genome-wide data sets [54]. Exploring commonalities and differences among fungi could be used to further understand the genetic basis for pathogenic phenotypes.

Table 1

Databases integrating host–pathogen data. All databases are publically available.

Database	URL	Description	Pathogen
BiologicalNetworks	http://flu.sdsc.edu/index.jsp	Interactive networks, proteomics, transcriptomics, and metabolomics	Viruses
GPS-Prot	http://www.gpsprot.org	HIV–human protein–protein interactions	Viruses
HPIDB	http://agbase.msstate.edu/hpi	Protein interaction database	Bacteria, viruses, fungi
PATRIC	http://www.patricbrc.org	Bacterial and human proteins	Bacteria
PHI-Base	http://www.phibase.org	Data-mining and gene expression data	Fungi, oomycetes, bacteria
PHIDIAS	http://www.phidias.us	Data-mining, genomic sequences and gene expression data	Bacteria, viruses, parasites, fungi
PhylomeDB	http://phylomedb.org	Gene phylogenies for phylogenetic and comparative analysis	Bacteria, fungi
PIG	http://molvis.vbi.vt.edu/pig	Protein interaction database	Bacteria, viruses, fungi
vHoT	http://dna.korea.ac.kr/vhot	Prediction of mRNA targets of viral microRNAs	Viruses

Infection modeling and microbial arcades

Spatio-temporal modeling of infection dynamics is an emerging field to incorporate the dynamics of pathogenesis [55,6]. One approach is evolutionary game theory, an application of game theory mathematics based on the relationship between the behavior of an organism and its evolution, or co-evolution, with other species. These studies formulate a simplified infection in silico and predict pathogen fitness by identifying game rules, often from genome-wide expression data. Most recently, it has been used to describe infections including: mixed viral infections of Arabidopsis thaliana [56], persistent bacterial infections [57], a simulated multi-species biofilm [58], and the mechanisms enabling survival of C. albicans inside macrophages [59]. For C. albicans, the outcome was analyzed based on the mean evolutionary cost of a cell population to obtain a positive fitness and the infection strategy employed by C. albicans to enable proliferation in the host was hypothesized be responsive to this cost. These studies emphasize the importance of analyzing microbes as adaptive social components of biological systems, because of their ability to sense and respond to the requirements of their own population, and that of their environment [60].

Computational modeling has been used to reconstruct the complex network between the immune cells and the bacterial pathogen Mycobacterium tubercolosis. On the basis of known interactions of the bacteria during infection, they estimated the influence of specific factors, such as an increase in specific cytokines or vaccination, on bacterial clearance and thereby identified the overall propensity for the bacteria to persist in the host under a wide range of conditions [61].

Most modeling approaches use genome-wide microarray expression data or RNA-seq. RNA-seq provides the advantage of simultaneous expression profiling of genes of the pathogens and their hosts, reducing concerns about platform-dependent effects. In addition, RNA-seq can potentially be used to investigate allelic variants of a transcript, and the evolution of microorganisms within its host. Small-scale network inference from the simultaneous analysis of C. albicans and DCs from M. musculus has predicted novel host–pathogen genetic interactions [62]. Furthermore, a genome-wide inference network using C. albicans has identified a number of candidate antifungal target genes [63]. These studies emphasize the advantages of simplifying genome-wide expression data using modeling and inference techniques to identify novel interactions and strategies utilized by the host and pathogen during infection.

Significant hurdles remain in order to use infection modeling on a large scale. One major limitation is that the experimental data is generated at different time scales. The transcriptional response of fungi takes place after minutes, proteomics from minutes to hours, and the subsequent immune response to the fungus from hours to days or even weeks. Choosing a mathematical approach to relate these time scales is not trivial. Moreover, the use of different units, strains, and animal models between laboratories can limit the ability to compare data sets. There been a push to standardize genome-wide data sets, including the Minimum Information for Biological and Biomedical Investigations (MIBBI, http://mibbi.sourceforge.net/), which will significantly aid in dataset comparison between laboratories. Relatedly, the maintenance and integration of new and existing fungal databases is needed to make the available information accessible and decrease the bottleneck for data analysis. Curation based on data models that incorporate pathway information [64] will make it easier to integrate new types of data sets, such as metabolomics, proteomics or host–pathogen data sets, as they become available.

Conclusions and outlook

A frequent critique of systems biology is that the massive influx of data has led to a fundamental loss of perspective because data generation has outpaced our capacity and ability to analyze them. It is therefore easy to loose the scale in which -omics data is biologically meaningful. Taking a lesson from Schrödinger's philosophy, the understanding of inner workings of the eye does not bring one closer to the perception of color: the additional information is irrelevant to the question. In other words, the biological context and proper parameter estimation of biological data sets is the key to generate models of predictive power. An initial definition of the system and its potential impact on the interacting species it contains is therefore required for analysis, including responses that determine pathogen clearance or host killing. Understanding the evolution of fungal strategies to survive and infect the host requires simultaneous investigation of microorganism–host interactions in both pathogenic and commensal species. Lessons learned from modeling the cell cycle show the importance of obtaining time course information either at the whole genome, or at the single molecule level, including the identification of biologically meaningful parameters, to obtain identifiable models. Developing strategies for the integration of multiple and complementary — quantitative -omics data sets, in a dynamic manner, will also be essential to further our understanding of microbial infections by reducing available data sets into testable models.

The host immune response is a complex entity and its behavior cannot be investigated in isolation from the environment that is driving adaptive changes, such as host immune defense. Systems biology holds the promise of helping us to obtain holistic views on the extent of this environment, and to generate predictions of host–microbe behavior and disease outcome. Combining the major schools of thought of mathematical modeling and functional genomics is a promising to solution to reach the goals of deciphering infectious processes and eventually improving therapeutic approaches to fungal infections.

Conflict of interest

The authors have declared that no conflicts of interest.

References and recommended reading

Papers of particular interest, published within the period of review, have been highlighted as:

• of special interest

•• of outstanding interest