SEARCH

SEARCH BY CITATION

Keywords:

  • fungi;
  • transcriptomics;
  • pathogenesis;
  • gene clusters;
  • telomeres

Abstract

  1. Top of page
  2. Abstract
  3. Introduction
  4. Models, methodologies and experimental designs
  5. Data similarities and divergences
  6. Locational aspects of gene expression and regulation
  7. Perspectives
  8. References

The capture of pathogen gene expression signatures directly from the host niche promises to fuel our understanding of the highly complex nature of microbial virulence. However, obtaining and interpreting biological information from infected tissues presents multiple experimental and intellectual challenges, from difficulties in extracting pathogen RNA and appropriate choice of experimental design, to interpretation of the resulting infection transcriptome, itself a product of responses to multiple host-derived cues. The recent publication of several host-infecting fungal transcriptomes offers new opportunities to study the commonalities of animal and plant pathogeneses, which in turn might direct the rational design of new and broader spectrum antifungal agents. Here, we examine the transcriptional basis of modelled Aspergillus fumigatus, Candida albicans, Cryptococcus neoformans, Ustilago maydis and Magneporthe infections, placing our analysis of the published findings within the context of the various modelling procedures used, and the relevant pathogen lifestyles, to facilitate the first cross-species comparison of fungal transcription during infectious growth. Significant concordance was identified among infecting transcriptomes of the inhaled fungal pathogens C. neoformans and A. fumigatus. The significance of gene clustering and subtelomeric gene repertoires is also discussed.


Introduction

  1. Top of page
  2. Abstract
  3. Introduction
  4. Models, methodologies and experimental designs
  5. Data similarities and divergences
  6. Locational aspects of gene expression and regulation
  7. Perspectives
  8. References

A fractional proportion of known fungal species is pathogenic. What distinguishes these virulent organisms from more than a million benign species is largely unknown; certainly their lifestyles and modes of pathogenesis are as varied as the range of diseases they cause. Despite this variance, commonalities at the molecular level are often found. Some regulatory pathways, for example nutrient acquisition, pH adaptation and morphogenetic reprogramming, are widely relevant to virulence in multiple species and hosts. However, neither aligned nor comparative transcriptional studies of disease-initiating fungi have been reported. Fungal genomic toolsets have recently benefitted from the publication of multiple pathogen genomes, and global transcriptional analyses are providing first insights on fungal gene expression during host colonization and invasion. Opportunities are therefore emerging to comparatively analyse host-invading fungal transcriptomes. In this minireview, we examine the results of recent investigations and ask whether it is possible to draw exploitable parallels or diversifications among the studies. We consider analyses of three human (Aspergillus fumigatus, Candida albicans and Cryptococcus neoformans) and two plant (Ustilago maydis and Magnaporthe grisea species complex) fungal pathogens (Table 1), giving careful consideration to methodological and technical limitations of the experimentation involved.

Table 1.   Organisms and studies analysed
Fungal species and strainsHostHost treatmentRoute of infectionOrgan/tissueTime points post infectionRNA amplificationExpression analysisReference sampleReferences
A. fumigatus Af293Male CD1 mice, 18–22 g, 4–5 weeks oldCyclophosphamide day −3, −1 and hydrocortisone acetate day −1Intranasal, 1 × 108 CFU per mouseLung14 hYesAmplicon arrayAf293, YPD, germlings developmentally matchedMcDonagh (2008)
C. albicans SC5314 and NGY152Male NZ white rabbits 2.5 ± 0.5 kgNoneMarginal ear vein 1–4 × 107 CFU kg−1KidneyDay +3NoEurogentec arraySC5314, RPMI 1640, mid-logWalker et al. (2009)
C. albicans SC5314 and ATCC10231Female BALB/c mice 8–10 week oldNoneIntraperitoneal 5 × 107 CFU per mouseLiver0, 0.5, 3, 5 hYesEurogentec arraySC5314, YPD, mid-logThewes et al. (2007)
C. neoformans H99Female A/Jcr miceNoneNasal inhalation 8.2 × 107 per mouseLung8 and 24 hNoSAGEC. neoformans H99, shift to low iron medium: 6 h growth, YNB broth: OD600 nm=14, rabbit CNSHu et al. (2008)
M. grisea KV1Rice (Oryza sativa) YT-16NoneRice sheath injection, 1 × 105 per sheathAdaxial epidermal layer and three mesophyll cell layers36 hYesAgilent array20% KV1 mycelia, with 80% gelatine infected host RNA, 3,3,3 medium, triply passaged for 14 hMosquera et al. (2009)
U. maydis FB1 and FB2Early Golden Bantam corn, 7 days oldNoneStem injection into the leaf whorl, suspension of OD600 nm=3Tumour formations on leaf bladesDay +13NoAffeymetrix arrayFB1, liquid array medium: A600 nm of 0.5Kamper et al. (2006)

Models, methodologies and experimental designs

  1. Top of page
  2. Abstract
  3. Introduction
  4. Models, methodologies and experimental designs
  5. Data similarities and divergences
  6. Locational aspects of gene expression and regulation
  7. Perspectives
  8. References

Six recent studies were included in our analysis. Methodological aspects (e.g. host species, immunosuppression and/or dosing regimens, etc.) of the reported experimentation are detailed in Table 1. Those that characterized host adaptation of the human respiratory pathogens A. fumigatus and C. neoformans (Hu et al., 2008; McDonagh et al., 2008) used mouse inhalational models of pulmonary infection, with subsequent bronchoalveolar lavage (BAL), to examine early-stage host adaptation in harvested fungal elements.

Aspergillus fumigatus is a common mould that causes opportunistic invasive infection in immunocompromised patients (Latge, 1999). To mimic this pathophysiology, mice were chemotherapeutically rendered neutropenic before infection. As A. fumigatus spores are abundant among the airborne microbial communities, and pulmonary infection is usually acquired following spore inhalation, mice were infected via the intranasal route (Fig. 1a), with a saline suspension of freshly harvested mitotic spores. Mice were culled at a time point (14 h) corresponding to the onset of pulmonary tissue invasion (Fig. 1b) and the transcriptome of infecting fungal germlings was analysed, relative to developmentally matched laboratory-cultured germlings, using doubly amplified mRNA populations.

image

Figure 1.  Modelling interactions between animal fungal pathogens and their experimental hosts (a) The intranasal inoculation route adopted for murine modelling of Aspergillus fumigatus infection (Mc Donagh et al., 2009) targets the pulmonary tract and cavity. (b) Representation of a neutropenic murine lung section during early-stage infection by A. fumigatus (14-h postinfection). The fungal burden increases as spores (in green) germinate, having sensed and adapted to the environmental challenges imposed by the host niche. (c) Schematic of a dysfunctional murine macrophage as it takes up A. fumigatus germlings and as it is later overcome by germinating spores. (d) The intranasal inoculation route adopted for murine modelling of Cryptococcus neoformans infection (Hu et al., 2008) targets the pulmonary tract and cavity. (e) Representation of a murine neutropenic lung section during early-stage infection of an immunocompetent host by C. neoformans. (f) Schematic of a murine macrophage as it takes up C. neoformans-encapsulated yeast cells (in green), and their re-emergence from the macrophage within several hours of phagocytosis (by analogy to in vitro data). (g) Diagram of the infection model used by Walker et al. (2009), a New Zealand white rabbit. Infection was intravenously introduced via the marginal ear vein, with the kidney being the target organ. (h) Cross-sectional view of a rabbit kidney showing Candida lesions, from which Candida albicans RNA was extracted. Lesions occur throughout the cortex. (i) Representation of C. albicans hyphal growth (green) showing the penetration of host cells and the disruption of the host tissue. (j) The murine infection model used by Thewes and colleagues used immunocompetent animals and an intraperitoneal route of infection. (k) Liver was the target organ, and surface adhesion and tissue penetration (l) were primary disease-initiating events captured during the timescale of transcriptional analysis.

Download figure to PowerPoint

A similar experimental protocol (Fig. 1d and e) was adopted by Hu and colleagues for C. neoformans, with the exception that mice were not immuncompromised, and two time points, 8 and 24 h, were adopted for the harvest of fungal elements. The serial analysis of gene expression (SAGE) methodology (Patino et al., 2002) was used to profile transcript populations from unamplified total RNA, obtained from the pooled contents of 20 murine BALs. SAGE ranks transcript abundance in RNA populations, which, with normalization between samples, can provide information on the relative transcript abundance between transcript populations. Therefore, no direct comparison with a reference sample was performed for this study; rather, the number of SAGE tags identified per transcript was recorded and tag populations were compared with those obtained in previous experimentation.

Various infection modelling options exist for Candida species as these organisms cause a range of infectious diseases. These vary in initiation site, and can invade multiple organs during systemic infection, although the kidney is the most common target in cases of murine intravenous infection (MacCallum et al., 2009). Systemic candidiasis is usually initiated when immunity is physically or chemotherapeutically impaired, and well-recognized risk factors for human systemic disease include catheterization, surgery and chemotherapy. Walker and colleagues studied the C. albicans transcriptome during rabbit renal infection (Walker et al., 2009), using an intravenous, ear vein infection (Fig. 2g) and a single 3-day time point. Fungal lesions (Fig. 1h and i) were harvested from the kidney with a scalpel and snap frozen before pooling, fixation and total RNA extraction. The large numbers of fungal cells obtained from these samples negated the requirement for mRNA amplification and the tissue fixation protocol was found to impact transcription minimally. The reference RNA sample was prepared from RPMI-cultured C. albicans cells (obtained from prior overnight culture in a rich medium and shifted for 6 h).

image

Figure 2.  Modelled interactions between fungal pathogens of plants and their host. (a) Following stem injection, the biotrophic plant pathogen Ustilago maydis initiates disease from a dikaryotic invasive filament and (b) proceeds via appressorium formation and tissue penetration through to (c) tumour formation. Tumour formation results from pathogen-induced plant growth alterations and the fungus proliferates and differentiates within the tumour tissue. Kamper et al. (2006) isolated total RNA from plant tumours at 13 days postinfection. Magnaporthe species undergo a series of morphogenetic transitions during the infection process. Following initial cutinase-mediated spore attachment to the rice leaf sheath (d), a narrow germ tube is generated, which differentiates into a penetrating appressorium (e), used by the fungus to gain entry into the leaf epidermis. Invasive hyphae are then generated that invaginate the plasma membrane to ultimately effect leaf lesions (f). Mosquera et al. (2009) targeted invasive hyphae as their sampled population.

Download figure to PowerPoint

Thewes and colleagues also studied systemic C. albicans infection, but in an immunocompetent murine host, analysing different phases of intraperitoneally administered infection, from liver attachment to penetration of liver surface-tissue, in time-course experimentation (Fig. 1j–l). In this instance, a YPD-grown comparator cell population was used for harvesting reference RNA (Thewes et al., 2007). RNA from infecting fungi was amplified before microarray hybridizations.

We selected two plant infection datasets. Kamper and colleagues analysed stem-injection-mediated infections of maize by the biotrophic plant pathogen U. maydis, initiating from a dikaryotic invasive filament and proceeding via appressorium formation and tissue penetration (Fig. 2a and b) through to tumour formation (Fig. 2c). During hyphal penetration, the host plasma membrane invaginates to form an interaction zone between the pathogen and the host (Fig. 2b). Tumour formation results from pathogen-induced plant growth alterations, with the fungus proliferating and differentiating within the tumour tissue. Kamper isolated total RNA from plant tumours at 13 days postinfection, providing sufficient RNA without amplication. The reference sample was cultured from one of the two infecting progenitors in minimal medium.

In the second plant infection study, Mosquera and colleagues studied the rice blast fungus Magnaporthe oryzae, a plant pathogen that threatens several agriculturally important food crops, predominantly rice (Wilson & Talbot, 2009). Magnaporthe oryzae undergoes a series of morphogenetic transitions during the infection process. Following initial cutinase-mediated spore attachment to the rice leaf sheath, a narrow germ tube is generated (Fig. 2d) that differentiates into a penetrating appressorium (Fig. 2e), used by the fungus to gain entry into the leaf epidermis. Invasive hyphae are then generated that invaginate the plasma membrane to ultimately effect leaf lesions (Fig. 2f). Mosquera and colleagues targeted invasive hyphae (Fig. 2f) as their sampled population in order to avoid filamentous necrotrophic hyphae characteristic of late-stage infection. Invading hyphae were harvested from leaf sheaths at 36 h postinfection, obtaining a relatively synchronous cell population in which most hyphae were inside first-invaded cells. Leaf sheaths were manually dissected in order to remove uninfected plant material and infected material was snap frozen before RNA extraction. RNA amplification was integral to the labelling protocol, with 500 ng of total RNA generating 10–15 μg of labelled cRNA.

All of the studies captured significant numbers of differentially expressed genes, where up/downregulated gene sets consisted of 1281/897 [9075] (McDonagh et al., 2008), 58/50 [85% of arrayed spots] (Walker et al., 2009), 255/221 [787] (Thewes et al., 2007); 1120/781 [15152] (Thewes et al., 2007) and 713/423 [6750] (Kamper et al., 2006), where square parentheses indicate the numbers of assayable spots per experiment. The C. neoformans SAGE analysis returned data on 84 gene tags (normalized to every 20 000 of the tag population sequenced), showing a higher representation relative to previously documented in vitro SAGE libraries, including a low-iron medium (LIM) SAGE library (Hu et al., 2007) against which most comparisons were made. We used several strategies to derive multispecies information on the co-ordinate regulation of homologous genes (Table 2) including best hit blast (Altschul et al., 1990) analysis, applied in a unidirectional sense, using peptides derived from the translation of species-specific differentially regulated transcript sequences. We also matched text descriptors from gene annotations in instances where spot annotations could not be readily matched to publicly accessible annotation databases or where significant redundancy of function among multiple gene identifiers might be expected (e.g. oxidoreductases and alcohol dehydrogenases).

Table 2.   Orthologous and analogous infection-related gene expression, among plant and mammalian fungal pathogen transcriptomes
 Hu (C. neoformans)McDonagh (A. fumigatus)Walker (C. albicans)Thewes (C. albicans)Kamper (U. maydis)Mosquera (M. grisea)
  1. We used several strategies to derive multispecies information on co-ordinate regulation of homologous genes, including best hit blast (Altschul et al., 1990) analysis, applied in a unidirectional sense, using peptides derived from translation of species-specific differentially regulated transcript sequences. We also matched text descriptors from gene annotations in instances where spot annotations could not be readily matched to publicly accessible annotation databases or where significant redundancy of function among multiple gene identifiers might be expected (e.g. oxidoreductases and alcohol dehydrogenases). Orange shading indicates upregulation, green shading indicates downregulation and grey shading indicates no change.

Glyoxylate cycle
 Acetyl-CoA synthetaseXM_567122AFUA_8g04770CA0848CA2858C46um133G_atAMG07946
 Malate synthaseXM_572462AFUA_6g03540CA4748CA4748W100um038G_at 
 Pyruvate decarboxylaseXM_567475AFUA_6g00750CA2474 C135um055G_atAMG01285
Tricarboxylic acid cycle
 AconitaseXM_570245AFUA_6g12930CA3546CA3546C45um172G_s_at 
 Succinate dehydrogenaseXM_572036AFUA_5g10370CA2470CA2470W60um108G_at 
 Malate dehydrogenaseXM_572038AFUA_2g13800CA5826CA5826C20um062G_atAMG08396.2/AMG14855.1
 Isocitrate dehydrogenaseXM_566837AFUA_3g15140CA4148CA4148/CA0643C125um139G_atAMG01022
Glycolysis
 Fructose-bisphosphate aldolaseXM_568771AFUA_5g10290CA5180CA5180C105um163G_atAMG12642
 PhosphofrucktokinaseXM_567487AFUA_4g00960CA3112CA3112 AMG13444.1
 HexokinaseXM_568853AFUA_2g16330CA0127CA0127C162um117G_at 
 EnolaseXM_569379AFUA_6g06770CA3874CA3874W20um076G_at 
Gluconeogenesis
 Phosphoenolpyruvate carboxykinaseXM_572603AFUA_6g07720CA3483CA5857C50um133G_atAMG12345.3
Other carbon metabolism genes
 Glutamine: fructose-6-phosphate amidotransferaseXM_570206 CA4016CA4016 AOS33118
 Glycogen phosphorylase-like proteinXM_571431AFUA_1g12920CA5206CA5206  
 d-Arabitinol dehydrogenaseXM_569729AFUA_5g08900CA3288CA3288C30um201G_at/C135um049G_atAMG13852
Lipid metabolism
 Butryrylcholinesterase triacyglcerol lipaseXM_567576AFUA_8g02530CA6071 W165um144G_at 
 Acyl-CoA oxidaseXM_568632AFUA_1G14850CA1572CA1572C15um226G_at 
 Enoyl-CoA hydratase/isomeraseXM_572730AFUA_2g10920CA5096CA5096C45um117G_at 
 2,4-Dienoyl-CoA reductaseXM_572896AFUA_6g11210CA3771 W25um185G_at/C25um199G_at/C50um162G_atAMG15192.1/AOS25954
 Enoyl-CoA hydrataseXM_566658AFUA_3g11480CA5060CA5060C45um117G_at 
 Long chain fatty acid CoA ligaseXM_572753AFUA_2g09910CA5992CA5992C45um129G_at 
Transport
 Monosaccharide transporterXM_568855AFUA_6g03060/AFUA_4g00800CA1506CA1506W45um126G_atAMG12940/AMG13082
 Plasma membrane iron permeaseXM_568258AFUA_5g03800CA5345CA5345W125um049G_at 
 High affinity copper transporterXM_570353AFUA_2g_03730CA1600CA1600W50um063G_atAMG06647.2
 GABA permeaseXM_568040AFUA_1g12310CA5039 W140um013G_atAMG07441\AMG00556.2
 Neutral amino acid permeaseXM_571440AFUA_8g02260CA1191 W20um225G_at 
 Sodium transporterXM_570596AFUA_4g09440/AFUA_6g03690CA4929CA4929W95um192G_atAOS20399
 Maltose permeaseXM_572882AFUA-2g10910CA1782CA1782C31um146G_atAMG15548
Virulence and stress response
 Heat shock protein 12XM_572088AFUA_6g12450CA4683CA4683W90um025G_at 
 Cu/Zn superoxide dismutaseXM_570285AFUA_5g09240CA4836CA4836W130um019G_at 
 Mannitol-1-phosphate dehydrogenaseXM_571772AFUA_7g01010CA4765CA2391C50um151G_atAMG15528
 Heat shock protein 90XM_568451AFUA_5g04170CA4959CA4959C190um019G_atAOS20508.2
 ThioredoxinXM_569667AFUA_8g01090CA6010CA6010C15um211G_atAOS25288\AMG06483
 Multicopper oxidase AFUA_4g14490CA1431CA1431C120um049G_atAMG05133/AOS32638/MGG_05790.6
 β-GlucosidaseXM_569544AFUA_6g08700CA5573 W165um003G_atAMG13799
 Carboxylesterase AFUA_8g00560CA1387  MGG_09081.5
 SubtilaseXM_567315AFUA_4g11800CA5322CA5322C80um258G_atAMG00014/MGG_10445.6
 Rare lipoprotein A AFUA_5g08030CA0689CA0689 MGG_07556.6
 Serine carboxypeptidaseXM_571636AFUA_5g07330CA5423CA542C149um157G_atMGG_03995.6
 Glycosyl hydrolaseXM_569830AFUA_2g17250CA1609CA1609W65um131G_atMGG_09030.6
 Fungal hydrophobin AFUA_6g06690CA2558CA2558W55um030G_atMGG_03671.6
 Serine carboxypeptidase AFUA_4g13750CA5476CA5476C149um157G_atMGG_10260.5

Data similarities and divergences

  1. Top of page
  2. Abstract
  3. Introduction
  4. Models, methodologies and experimental designs
  5. Data similarities and divergences
  6. Locational aspects of gene expression and regulation
  7. Perspectives
  8. References

Despite the variance among the size of datasets and disparate infection models, some interesting observations can be drawn from the comparison. We found impressive concordance between upregulated A. fumigatus and C. neoformans genes (Table 2). Such a similarity of the transcription profile is even more remarkable, given the varying immunosuppressive regimens used and different morphogenetic programmes of the two species (yeast vs. filamentous fungus). This intriguing finding may therefore reflect the similarity of nutrient sources and physiological conditions (such as temperature, iron limitation and oxygen tension) in the mammalian pulmonary niche and the dominance of such factors over immune status and species-specific traits. Despite the similarities in infection modelling procedures, the progression of infection would have differed significantly between the McDonagh and Hu studies in respect of the differential pathogenic strategy adopted by A. fumigatus and C. neoformans, respectively. Murine innate defence against A. fumigatus is sufficient to prevent infection, even in heavily infected animals, and immunosuppression is required to establish infection (Lewis & Wiederhold, 2005). In the McDonagh study, both macrophage and neutrophil cell populations were chemotherapeutically targeted, using hydrocortisone acetate and cyclophosphamide, respectively, the former drug administered in a single dose 1 day before infection and the latter periodically administered throughout the duration of the experiment (Lewis & Wiederhold, 2005). Phagocytosis by macrophages harvested from hydrocortisone-treated A. fumigatus-infected mice is known to occur, but fungal killing is compromised (Philippe et al., 2003). It is likely, therefore, that host cells, predominantly macrophages, are encountered in the alveolar and bronchial spaces (Fig. 2b and c), and encountered macrophages are compromised in their ability to kill A. fumigatus spores. Conversely, the encapsulated facultative intracellular pathogen C. neoformans can establish infection in immunocompetent mice. Moreover, the interaction between macrophages and C. neoformans is critical for containing the dissemination of this pathogenic yeast, whose success is subverted by C. neoformans-derived factors. Cryptococcus neoformans is capable of replication within the macrophage phagolysosome, a process that ultimately leads to host cell lysis or phagosome extrusion (Tucker & Casadevall, 2002; Alvarez & Casadevall, 2006; Ma et al., 2006). As in vitro studies indicate that the time taken to extrude a C. neoformans-containing phagolysosome can be as little as 2 h (Tucker & Casadevall, 2002; Alvarez & Casadevall, 2006; Ma et al., 2006), it is likely that multiple macrophage encounters occurred during the experimental time frame, and, contrary to the A. fumigatus infection model, noninfected macrophages were completely proficient with respect to killing ability.

Carbon metabolism was, to varying degrees, commonly implicated among all of the mammalian pathogen datasets with acetyl-CoA synthetase and isocitrate dehydrogenase featuring in all four upregulated genesets. Combined with extant data on fungal carbon-metabolizing gene products and virulence, considerable insight can be gained from our comparative analysis. Firstly, the differential roles of glyoxylate cycle enzymes in virulence, which has been studied in multiple mammalian fungal pathogens, could not have been predicted from our comparative transcriptomic analysis. Glyoxylate cycle gene products are required for full virulence in C. albicans (Lorenz & Fink, 2001; Wang et al., 2003; Barelle et al., 2006) and M. grisea (Wang et al., 2003), but not in A. fumigatus (Schobel et al., 2007; Olivas et al., 2008) or C. neoformans (Rude et al., 2002). Indeed, based on our analysis, one might have predicted the necessity of glyoxylate pathway functionality in C. neoformans and A. fumigatus and nonrequirement in M. grisea (Table 2). Thus, the first of several pitfalls inherent to interpretation of these analyses becomes apparent.

Interpretation of data comparisons, and subsequent predictions of virulence genes, are heavily dependent on the experimental design, and relate directly to the choice of the time point(s), choice of the reference sample(s) and reliance on data drawn from populations of cells. Single time-point analyses evidently do not provide the resolving power necessary to predict virulence determinants relevant to multistage pathogenetic processes, as evidenced by the requirement for glyoxylate cycle-encoding gene products, acting at prepenetrative stages of infection, for virulence in M. grisea (Wang et al., 2003) and their apparently static levels of transcription (Table 2) in invasive hyphae. For comparative microarray analyses (including the choice of the comparator SAGE tag library in SAGE analytical approaches), the origin of the reference sample profoundly impacts on up- and downregulated genesets. It may, therefore, be naive to expect experiments using reference samples of diverse nutrient compositions (e.g. YPD, RPMI1640 and LIM) to result in similar gene expression profiles. A case in point is provided by a collective impediment to fungal propagation in plant and animals: the lack of available iron, which is an essential cofactor for many cellular processes. Ustilago maydis, M. grisea and A. fumigatus use siderophores, a class of nonribosomal peptide synthase (NRPS)-dependent secondary metabolites, to scavenge ferric ion selectively through the formation of soluble chelation complexes (Schrettl et al., 2007; Bolker et al., 2008; Hof et al., 2009). Intra- and extracellular siderophores are required for full virulence in a pulmonary murine model of invasive aspergillosis (Schrettl et al., 2007), and accordingly, gene expression at siderophore biosynthetic gene clusters was induced in a similar murine model at 14-h postinfection, indicating that the response to iron limitation in the mammalian host is addressed at a very early stage of infection (McDonagh et al., 2008). Therefore, concordance between transcriptional data and important virulence determinants can be expected from this type of analysis. However, despite the observed similarity of gene expression profiles between A. fumigatus and C. neoformans, iron acquisition was not identified as an important component of the infecting C. neoformans transcriptome. This may, in part, be due to the use of an LIM comparator in the C. neoformans experimentation, which would undoubtedly occlude, at the transcriptional level, this aspect of pathogenic growth. While C. neoformans does not synthesize siderophores, iron acquisition is crucial for C. neoformans virulence (Jung et al., 2009). Thewes and colleagues also found gene expression that reflected iron limitation.

The differential choice of reference samples for the Walker and Thewes studies may, notwithstanding different modelling procedures, account for the vast differences in gene expression profiles noted between the two studies (Table 2) of this mammalian pathogen. Moreover, although the poor concordance between previously identified virulence factors (based on murine experimentation) and differentially regulated genes is noted by the authors of Walker et al, it is not possible to comment upon the relevance of this observation, given the absence of virulence data in the rabbit model of infection and the differing scale of experimentation.

We found little concordance between metabolic functions upregulated in animal vs. plant pathogens, an observation that may have relevance to the differential retention of saprophyte gene sets among plant pathogens. Similarities, where found, reside in transport, virulence and stress-related functional cohorts (Table 2). Moreover, a striking similarity in higher order gene regulatory activity can be found in instances where positional information is easily retrievable from genome annotation. Thus far, the phenomenon has been reported in U. maydis (Kamper et al., 2006), A. fumigatus (McDonagh et al., 2008) and M. grisea (Collemare et al., 2008), although few microarray datasets have been appropriately scrutinized.

Locational aspects of gene expression and regulation

  1. Top of page
  2. Abstract
  3. Introduction
  4. Models, methodologies and experimental designs
  5. Data similarities and divergences
  6. Locational aspects of gene expression and regulation
  7. Perspectives
  8. References

A significant paradigm shift in eukaryotic genome biology was the discovery that genes involved in functionally related pathways often cluster at proximal genomic locations (Keller & Hohn, 1997). The sequencing of numerous pathogen genomes and advances in bioinformatic and molecular biology has reinforced gene clusters as a common feature of fungal genomes. The term ‘cluster’ has been used to refer to significant portions of DNA enriched for certain features, such as transposons located in centromeric regions of the C. neoformans genome (Loftus et al., 2005), or lineage-specific genes found in 13 chromosomal islands of the A. fumigatus genome (Fedorova et al., 2008). The term is also used to refer to smaller numbers of genes located adjacently within relatively small loci. Such contiguous genes can collectively direct the biosynthesis of a small molecule, such as a secondary metabolite (Keller et al., 2005), or may simply be genes of related function, such as clusters of genes with putative signal peptides found in U. maydis (Kamper et al., 2006). The size, gene content and products of clusters are diverse; of special interest to the study of pathogenesis is the enrichment of virulence-associated genes within large chromosomal regions or their presence in contiguous clusters. These phenomena pose two challenging questions: what is the impact of the encoded biosynthetic products during pathogenesis and why are some virulence-associated genes clustered? In vivo gene expression profiling of clinically and agriculturally relevant fungal pathogens is proving to be a highly useful tool for determining the evolutionary origin of clusters and their impact on virulence.

A subset of clusters putatively involved in the virulence of important pathogens produce secondary metabolites. A key enzyme, most commonly an NRPS or polyketide synthase, produces a precursor molecule, which is subsequently modified by other enzymes encoded by the cluster. These genes usually produce a single product of small molecular weight, for example polyketides (lovastatin and aflatoxin B1), nonribosomal peptides (penicillin G and gliotoxin), terpenes (gibbererellin) and indole alkaloids (fumitromorgin C), which are dispensable for cellular growth and have a restricted taxanomic distribution (Keller et al., 2005). The well-documented cytotoxic and phytotoxic properties of many of these compounds have long identified them as putative virulence factors. Gene expression profiling and candidate gene analysis of multiple secondary metabolite-producing species present us with the first opportunity to assess their role as a common molecular feature used by fungi to overcome universal challenges encountered in the host niche. Other secondary metabolites have impacts on virulence that are unique to both the host environment and the stage of infection. The immunotoxic dipeptide gliotoxin, produced by a cluster of 19 genes in A. fumigatus (Cramer et al., 2006), is induced 14-h postinfection relative to laboratory culture (McDonagh et al., 2008) and is a virulence factor in a hydrocortisone acetate-treated, but not neutropenic, murine infection model. The action of gliotoxin as a virulence factor in vivo is most likely due to action against neutrophils, which is supported by ex vivo cellular assays (Bok et al., 2006a), and has recently been substantiated as acting at the level of proapoptotic gene family members in a physiologically relevant context using hydrocortisone acetate-treated BAK knockout mice (Pardo et al., 2006).

A unifying model to explain the existence of fungal clusters is currently unavailable, but it seems clear that multiple evolutionary mechanisms may explain their origin. Currently, three main hypotheses have been proposed, and gene expression analysis represents a useful tool for determining the relative contribution of these hypotheses to fungal gene clustering. Horizontal gene transfer (HGT) suggests that clusters may both originate and be maintained from selective pressure following HGT events. Gene clusters that encode an entire metabolic pathway or virulence factor are more likely to result in a phenotypic advantage to the recipient genome (Walton, 2000). The gene duplication, diversification and differential gene loss (DDL) hypothesis emphasizes the fundamental features of specific genomic regions as being the driving force behind clustering. Areas of microbial eukaryotes that are most subject to high genetic and genomic variability are the subtelomeres, located between the telomeric end of linear chromosomes and chromosome-specific sequences (Farman, 2007). These regions are characterized by repeat sequences of various lengths, enabling high levels of ectopic recombination, a phenomenon facilitated by telomere grouping at the nuclear periphery (Barry et al., 2003). Additionally, many subtelomeres are enriched with retrotransposons and other mobile genetic elements, which can lead to local insertions, deletions, duplications and inversions (Volff, 2006). Virulence-associated genes are often located in pathogen subtelomeres, and include those directing antigenic variation in Plasmodium falciparum and Trypanosoma brucei, surface glycoproteins of Candida glabrata (De Las et al., 2003), secondary metabolites, catabolism and transport in A. fumigatus (Fedorova et al., 2008) and secondary metabolites in U. maydis (Bolker et al., 2008). While targeted sequencing of M. grisea chromosomes demonstrated no virulence-associated gene enrichment at subtelomeres (Farman, 2007), gene expression analysis of A. fumigatus germlings during host invasion found that genes induced during infection displayed subtelomeric and lineage-specific bias, supporting the diversity of these regions being more important than HGT in the evolution of pathogenicity for this species (McDonagh et al., 2008). Both the HGT and DDL hypotheses suggest that an increase in the virulence-associated gene content at restricted genomic locations leads to an increase in the pathogenicity spectrum within a population, enabling differential survival in the host niche. A different hypothesis suggests that clustering may facilitate the epigenetic regulation of functionally related genes during niche adaptation. This model stresses that genes grouped in close proximity may be regulated by gross modifications to the chromosome environment, such as the boundaries between euchromatin and heterochromatin. This model is contingent with the discovery of an Aspergillus methyl transferase, LaeA, which was found to be required for the production of many secondary metabolite toxins and essential for virulence in a murine model of infection (Bok et al., 2005). The movement of genes into or out of these clusters leads to gain or loss of LaeA regulation, respectively, suggesting that this global regulator of secondary metabolite biosynthesis regulates gene expression at the level of chromatin remodelling (Bok et al., 2006b). In vitro gene expression analysis shows that LaeA regulates the expression of key genes at multiple secondary metabolite loci, including gliotoxin and the cytotoxic quinine pseurotin (Perrin et al., 2007). These data collectively support the hypothesis that clusters facilitate epigenetic control of functionally related genes that are required for virulence. Of great interest will be the global expression analysis of the ΔlaeA strain during infection, to determine the epigenetic regulation of virulence-associated clusters in vivo (Cairns et al., unpublished data). Other pathogens display remarkable coordination of cluster-related gene expression during infection. For example, 12 U. maydis clusters, harbouring genes encoding secreted proteins, display similar expression patterns during infection; five of these clusters are essential for full virulence (Kamper et al., 2006). It is interesting to speculate that epigenetic factors may both control the expression and contribute to the maintenance of clusters in pathogens of animals and plants. The presence of virulence genes within clusters has prompted comparisons with the prokaryotic pathogenicity island phenomenon (Dean, 2007). Whether the molecular basis of fungal virulence will be as drastically altered by the discovery of pathogenicity clusters remains to be seen. What is clear is that gene expression analysis of multiple pathogens during infection has contributed considerably to our understanding of the role and evolutionary origins of these intriguing genomic attributes.

Perspectives

  1. Top of page
  2. Abstract
  3. Introduction
  4. Models, methodologies and experimental designs
  5. Data similarities and divergences
  6. Locational aspects of gene expression and regulation
  7. Perspectives
  8. References

Clearly, there is much to be gained from comparative analysis of fungal transcriptomes during the initiation of infection. In addition to the pitfalls introduced by experimental design considerations, the overriding obstruction encountered during our comparative analysis was the impenetrable nature of the published genesets, genome databases and comparative genomics tools. Although the advent of postgenomic fungal analyses has prompted investment in supportive bioinformatic tools, a one-stop comparative genome database that relates directly to gene product function, homologues in other fungi, genome location, spot positions on microarrays and representation in other datasets does not exist for any fungal pathogen (although we are currently developing such tools for A. fumigatus). Analyses such as ours, therefore, take many months to perform, constitute publishable studies in themselves and remain relatively primitive with respect to the accuracy of homologue predictions. Such shortcomings must be addressed if the full benefit of comparative studies is ever to be realized within a practicable timescale for a single researcher. This requires appropriately formatted datasets and databases that interconnect data of diverse species origins, a goal that must now become a priority if resources and generated experimental data are to be maximally exploited.

References

  1. Top of page
  2. Abstract
  3. Introduction
  4. Models, methodologies and experimental designs
  5. Data similarities and divergences
  6. Locational aspects of gene expression and regulation
  7. Perspectives
  8. References