Characterisation of the cellular and proteomic response of Galleria mellonella larvae to the development of invasive aspergillosis

Background Galleria mellonella larvae were infected with conidia of Aspergillus fumigatus and the cellular and humoral immune responses of larvae to the pathogen were characterized as invasive aspergillosis developed. Results At 2 h post-infection there was an increase in hemocyte density to 7.43 ± 0.50 × 106/ml from 0.98 ± 0.08 × 106/ml at 0 h. Hemocytes from larvae immune primed for 6 h with heat killed A. fumigatus conidia displayed superior anti-fungal activity. Examination of the spread of the fungus by Cryo-imaging and fluorescent microscopy revealed dissemination of the fungus through the larvae by 6 h and the formation of distinct nodules in tissue. By 24 h a range of nodules were visible at the site of infection and at sites distant from that indicating invasion of tissue. Proteomic analysis of larvae infected with viable conidia for 6 h demonstrated an increase in the abundance of gustatory receptor candidate 25 (37 fold), gloverin-like protein (14 fold), cecropin-A (11 fold). At 24 h post-infection gustatory receptor candidate 25 (126 fold), moricin-like peptide D (33 fold) and muscle protein 20-like protein (12 fold) were increased in abundance. Proteins decreased in abundance included fibrohexamerin (13 fold) and dimeric dihydrodiol dehydrogenase (8 fold). Conclusion The results presented here indicate that G. mellonella larvae may be a convenient model for studying the stages in the development of invasive aspergillosis and may offer an insight into this process in mammals. Electronic supplementary material The online version of this article (10.1186/s12866-018-1208-6) contains supplementary material, which is available to authorized users.


Background
Galleria mellonella larvae are an ideal in vivo model to quickly and easily assess the virulence of a range of human pathogens, to comprehensively analyze the hostpathogen interactome and to evaluate the in vivo toxicity and efficacy of antimicrobial agents [1][2][3]. G. mellonella larvae are easy to inoculate, generate results within 48 h and are free from the ethical and legal restrictions which surround the use of mammals for this type of testing [4,5]. Microbial virulence and pathogenesis can be assessed using a variety of endpoints including survival, fluctuations in hemocyte density, oscillations in fungal burden and changes in hemolymph proteome [6,7]. Insect hemocytes can be easily isolated and used in ex vivo cellular assays to determine phagocytepathogen interactions [8].
G. mellonella larvae have been used to assess the virulence of a range of fungal pathogens including Aspergillus fumigatus, Candida albicans and Cryptococcus neoformans and generated results with strong correlations to those established in mammals [9][10][11]. The generation of comparable results is due to many similarities which exist between the mammalian innate immune response and the insect immune system [5,12]. Insect hemocytes have receptors (Toll, β-1,3-glucan, etc) and signaling pathways (IMD, JNK, JAK/STAT) similar to those of mammalian neutrophils [12,13]. Hemocytes can engage in lectin mediated phagocytosis of opsonized microbial cells. Hemocytes can also incorporate proteins homologous to neutrophil p40 phox , p47 phox , p67 phox and gp91 phox into their membranes which contribute to the formation of an active NADPH oxidase complex capable of producing superoxide [8,14]. Hemocytes also undergo degranulation and release a range of granular proteins into the phagosome [12]. Extracellular killing is mediated by degranulation, the formation of NET-like structures and the formation of macro-cellular structures (melanotic encapsulation and nodulation) homologous to mammalian complement proteins and granuloma formation [8,[15][16][17][18]. Furthermore, A. fumigatus toxins gliotoxin and fumagillin inhibit the microbicidal activity of human neutrophils and insect hemocytes by blocking the formation of F-actin [19,20].
A. fumigatus can induce allergic, saprophytic and invasive disease depending upon the host immune status. In the case of invasive disease inhaled conidia germinate, form hyphae within alveoli and migrate beyond the pulmonary epithelium and into the bloodstream (angioinvasive), ultimately disseminating throughout the host if anti-fungal therapy is not commenced [21,22]. In the murine model of chronic granulomatous disease (CGD), invading hyphae are found within granulomatous lesions predominated by neutrophils that probably function to prevent spread of conidia [23]. Larvae of G. mellonella have been utilized to study the virulence of A. fumigatus and provided results comparable to those from murine studies. A. fumigatus Δpes3 mutants displayed significantly higher mortality in G. mellonella than WT A. fumigatus which was mirrored in two murine models of pulmonary aspergillosis (corticosteroid-treated and neutropenic) [24]. Mutant strains of A. fumigatus (ΔsidA, ΔsidC, ΔsidD, ΔsidF, Δpaba and ΔcpcA) demonstrated almost complete correlation of virulence when assessed in G. mellonella and mice [9]. Hemocytes can discriminate between non-germinated, germinating and hyphal forms of A. fumigatus which is also the case with human neutrophils [25,26]. Aspergillus terreus infection in G. mellonella demonstrated unique histological findings consistent with those observed in disseminated aspergillosis in mammals [27].
The aim of the work presented here was to analyze the response of G. mellonella larvae to A. fumigatus infection and to examine similarities with invasive aspergillosis in mammals. While G. mellonella larvae are now widely used as in vivo models a greater understanding of the pathogenhost interaction may allow the identification of a range of end points and further validate use of larvae for studying disease processes in vivo.

Modeling invasive and disseminated aspergillosis in G. mellonella
Cryo-imaging was employed to visualize the stages of invasive A. fumigatus disease in G. mellonella larvae. Larvae were inoculated with viable A. fumigatus conidia through the last left proleg in the posterior region (see white-edged arrow). Small discrete nodules appeared in the anterior region and around the perimeter of the haemocoel 6 h post infection (see black arrows) indicating dissemination of the A. fumigatus conidia from the site of infection. By 24 h there is extensive melanization of larval tissue and cuticle (see white arrows) indicating invasion from the insect haemocoel into surrounding tissue, and formation of both large diffuse fungal nodules at the site of inoculation and smaller distinct fungal nodules throughout the larva (Fig. 4).
Melanized lesions from larvae infected with A. fumigatus for 6 and 24 h were extracted and dissected. Dense hemocyte infiltration (green arrow) and melanized plaques (gray arrow) were visible by light microscope (Fig. 5). Confocal laser scanning microscopy of nodules isolated at 6 and 24 h and analysis of fluorescence due to calcofluor white binding confirmed the presence of germinated conidia (germ tube; black arrow) at 6 h and dense hyphal infiltration (white arrow) at 24 h post infection (Additional file 1: Figure S1). Interestingly, the use of fluorescence microscopy at 6 and 24 h revealed the presence of germinated conidia (white arrow) and hyphae (red   Figure S2) and reported elsewhere [28]. Conidia and hyphae remained viable inside granulomatous tissue as determined via plate assay.  Label free quantitative proteomic analysis was performed on the G. mellonella cell free hemolymph after exposure of larvae to viable A. fumigatus conidia (1 × 10 6 / larva) for 0, 6 and 24 h. In total, 2517 peptides were identified representing 252 proteins with two or more peptides and 28 and 108 (6 h v 0 h and 24 h v 0 h, respectively) proteins determined to be differentially abundant (ANOVA, p < 0.05) with a fold change of > 2. A total of 17 proteins were deemed exclusive (i.e. with LFQ intensities present in all four replicates of one treatment and absent in all four replicates of the other two treatments). These proteins were also used in statistical analysis of the total differentially expressed group following imputation of the zero values as described. After data imputation these proteins were included in subsequent statistical analysis. A principal component analysis (PCA) performed on all filtered proteins distinguished the 0, 6 and 24 h A. fumigatus treated samples indicating a clear difference between each proteome (Additional file 3: Figure S3A). Hierarchical clustering of z-score normalized intensity values for all significantly differentially abundant proteins (n = 252) produced the three replicates of each sample group (Additional file 3: Figure S3B). Furthermore, 2 major protein clusters were identified: 0 h and 6 h abundant proteins (Cluster A) and 24 h abundant proteins (Cluster B).

Discussion
G. mellonella larvae are widely utilized in determining in vivo antimicrobial drug activity, in assessing relative toxicity of compounds and in quantifying the virulence of bacterial and fungal pathogens [2,[29][30][31]. While G. mellonella larvae offer many advantages as in vivo models, there is the possibility of extending their use for modeling disease processes [2,3]. This creates an opportunity to model infection processes in larvae that show similarities to those in mammals and to design therapies in the insect system that may be translated into mammals. The results presented here characterize the immune response of G. mellonella to A. fumigatus and describe the dissemination of the fungus through the larva in the first 24 h of infection. Inoculation of viable conidia (1 × 10 6 / larva) resulted in a significant increase in the density of circulating hemocytes at 2 h as compared to the PBS control possibly due to the release of sessile hemocytes normally attached to the inner surface of the hemocoel or fat body of larvae [32]. Heat killed conidia produced a greater increase in hemocyte density and it is possible that loss of the rodlet/hydrophobin layer on their surface may trigger a larger immune response [33]. Heat killed conidia and germinating conidia display β-glucan on their cell surface and recruit neutrophils into the airways of C57BL/6 mice [34]. Hemocyte density was significantly elevated for 6, 8 and 10 h post infection with viable conidia and this may be associated with the formation of hyphae and secretion of metabolites from A. fumigatus. Hemocyte populations from larvae inoculated with heat killed A. fumigatus for 6 h displayed increased fungicidal activity as compared to hemocyte populations from control and 24 h treated larvae.
In mice, A. fumigatus increased the activity of both macrophages and neutrophils, which are essential in conidia and hyphal destruction by phagolysosome mediated killing, degranulation and NET formation [35,36]. Interestingly, mice exposed to conidia show a significant increase in total immune cell density (in bronchoalveolar lavage fluid) but specifically an increase in macrophages after 4 h and neutrophils 48 h post exposure [37]. Histology of the CGD invasive aspergillosis mouse lung shows germinated conidia early and the formation of hyphae after 24 h with neutrophil infiltration and pyogranulomatous lesions surrounded by granulocytes [23].
Cryoviz imaging and fluorescent microscopy characterized the dissemination and development of Aspergillus within larvae. By 6 h, fungal infection had spread to distal sites of the larvae and is marked by the formation of well-defined melanized nodules consisting of granulocyte infiltration and encapsulation of A. fumigatus germinating conidia. Here, hemocytes accumulate around antigenic A. fumigatus conidia preventing its spread. Hemocyte membrane was found surrounding A. fumigatus conidia and hyphae which is consistent with encapsulation responses of G. mellonella larvae to foreign material [18]. Larvae at twenty four hours post infection were characterized by melanization of the cuticle, the appearance of de novo nodules and expansion of fungal nodules from the site of inoculation throughout the host. These nodules show similarities to granulomatous structures which are characteristic of invasive aspergillosis in the CGD murine model. Histology from the CGD murine lung illustrates granulomatous lesions with significant hyphal invasion [23]. However, insect granulomatous structures lack lymphocytes, key components of mammalian granulomas. A. fumigatus remained viable and grew through melanized nodules 6 and 24 h post infection thus highlighting the importance of nodulation as an early response to an invading fungal inoculum.
Quantitative proteomics was employed to characterize the larval humoral immune response to A. fumigatus. At 6 h there was an increase in the abundance of antimicrobial peptides and proteins (gloverin, moricin, lysozyme, cecropin) and proteins associated with the prophenoloxidase cascade (serpin-4B, prophenoloxidase activating enzyme 3 and prophenoloxidase activating factor 3) in hemolymph. During the mammalian innate response to A. fumigatus, a range of antimicrobial peptides (defensins, cathelicidins) and proteins (lactoferrin, lysozyme) are produced and these are essential in curtailing early fungal establishment and growth [38,39]. Furthermore, the prophenoloxidase cascade is analogous to the mammalian complement protein cascade in terms of protein structure, function and mode of action [40,41].
At 24 h post infection antimicrobial peptides and prophenoloxidase family members are increased but also proteins associated with tissue invasion (muscle protein 20 like protein), recognition and opsonization of fungal cells (hemolin, peptidoglycan recognition like protein B) and inhibition of fungal proteinases (insect metalloproteinase inhibitor (IMPI)) [42,43]. IMPI is induced following fungal infection in G. mellonella and functions to inhibit the activity of secreted metalloproteinases which act as virulence factors to degrade host defenses [44]. A. fumigatus produces a variety of metalloproteinases most notably is Asp f5/mep, a 42 kDa Zn/Mep which possesses collagenlytic and elastinolytic activity and is important for immune cell recruitment in the murine lung [45,46]. Gustatory protein was increased in abundance at 6 and 24 h post A. fumigatus infection and was also increased in G. mellonella larvae in response to entomopathogenic fungal culture filtrate and is hypothesized to be associated with altered feeding responses and possibly toxin avoidance in insects [47,48]. Hdd1 and Hdd11 were found increased in 24 h A. fumigatus infected larval hemolymph. Both are induced following bacterial infection in Hyphantria cunea with the former sharing homology with mucin-5 AC-like protein from Plutella xylostella and the later homologous with Noduler from Antheraea mylitta [49,50]. Noduler shares a reeler domain with Hdd11 and binds both insect hemocytes and fungal β-1, 3 glucan, is enriched in nodules and may act as a facilitator of nodulation [51]. Interestingly, 27 kDa hemolymph protein, Lysozyme-like protein 1, C-type lectin 10 precursor and β-glucan binding protein were significantly decreased at 24 h possibly as a result of binding to hyphal structures and facilitating hemocyte recognition of fungal structures. Anionic antimicrobial peptide 2 was decreased in abundance 24 h after A. fumigatus infection (Fig. 6b) and was also decreased in G. mellonellla infected with C. albicans [52]. At 6 h post infection there is a decrease in Apolipophorin III in hemolymph. Apolipophorin III mediates recognition of fungal conidia, activates the phenoloxidase cascade and dose dependently enhances melanized nodule formation [53,54]. Its decreased abundance in hemolymph may be as a result of increased binding to A. fumigatus at the site of nodule formation [55].

Conclusion
Due to the lack of a comparable respiratory system to mammals, the use of G. mellonella to study A. fumigatus interactions at the broncoalveolar surface is not possible. However, the results presented here, reveal similarities between the development of invasive aspergillosis in mammals and in G. mellonella larvae. This work has characterized the cellular and humoral immune responses of G. mellonella larvae to A. fumigatus as well as the morphological changes of conidia and dissemination of hyphae throughout larvae. These results document significant parallels between the mammalian and insect responses to A. fumigatus infection, and illustrate how the development of invasive aspergillosis in larvae shows similarities to that which occurs in mammals. Studying the development of invasive aspergillosis in larvae may give novel insights into the pathogenhost interactions that could improve our understanding of this disease process in humans.

A. fumigatus culture conditions
A. fumigatus ATCC 26933 (obtained from the American Type Culture Collection) was used in this study. Aspergillus cultures were grown in sabouraud dextrose broth (SAB) (Sigma Aldrich) at 37°C and 200 rpm for up to 2 days. Stocks were maintained on malt extract agar (MEA) (Oxoid).

Larval culture and inoculation
Sixth instar larvae of the greater wax-moth G. mellonella (Livefoods Direct Ltd., Sheffield, England), were stored in the dark at 15°C to prevent pupation. Larvae weighing 0.22 ± 0.03 g were selected and used within 2 weeks of receipt. Ten healthy larvae per treatment and controls (20 μl PBS for appropriate incubation time); (n = 3) were placed in sterile 9 cm Petri dishes lined with Whatman filter paper and containing wood shavings. Larvae were inoculated with viable or non-viable (heat-killed) conidia through the last left pro-leg into the hemocoel with a Myjector U-100 insulin syringe (Terumo Europe N.V., Belgium). Larvae were acclimatized to 37°C for 1 h prior to all experiments and incubated at 37°C for all studies. All experiments were performed independently on three separate occasions.

Determination of hemocyte density
Larvae were inoculated with 20 μl of phosphate-buffered saline (PBS) containing heat killed, PBS containing viable A. fumigatus (1 × 10 6 ) conidia or PBS. Heat killed (HK) conidia were produced by heating to 95°C for 20 min and plated on SAB agar to confirm loss of viability. Changes in hemocyte density were assessed by bleeding 40 μl each from 5 G. mellonella larvae into a micro-centrifuge tube on ice, to prevent melanization. Hemolymph was diluted in 0.37% (v/v) mercaptoethanol supplemented PBS and cell density was determined using a hemocytometer. Cell density was expressed in terms of hemocytes per ml of hemolymph. Statistical analysis was performed by comparing treatments to PBS injected controls at respective time points. Experiments were performed on three independent occasions and the means ± S.E. were determined.

Determination of fungicidal activity of hemocytes
Larvae were inoculated with 1 × 10 6 heat killed conidia and incubated for 0, 6 or 24 h (n = 10) at 37°C. Hemocytes were extracted, washed and mixed with cell free hemolymph opsonized C. albicans (2 × 10 6 cells) at 37°C at a 2:1 ratio (hemocytes: yeast cells) in PBS as previously described [8]. Aliquots were taken at 0, 20, 40, 60 and 80 min, diluted and plated onto YEPD agar plates. Colonies were counted and expressed in terms of percentage of the original number of cells at time zero. Results were calculated as the mean (± S.E.) from at least three experiments with colony counts performed in triplicate for each sample.
Cryo-imaging of A. fumigatus infection in G. mellonella G. mellonella were inoculated with 1 × 10 6 viable A. fumigatus conidia for 0, 6 and 24 h. Larvae were placed on ice to inhibit any movement. They were embedded in Bioinvision Cryo-Imaging Embedding Compound and flash-frozen in liquid nitrogen and later mounted on a stage for sectioning. Sectioning and imaging was carried out every 10 μm using a Cryoviz™ (Bioinvision Inc., Cleveland, OH) cryo-imaging system.

Fluorescent imaging of fungal nodules
G. mellonella larvae infected with A. fumigatus for 6 and 24 h were dissected in PBS and nodules dissected apart with fine needles, transferred to a glass slide, washed with PBS 3 times and stained with Calcofluor white and FM 4-64 (Sigma) for 30 min at 16°C. The cells were washed twice (PBS) and a cover slide was placed on top. The cover slides were then fixed in-situ by applying a clear sealing solution around the perimeter of the slide which also prevented the sample drying out. Cells were viewed with an Olympus Fluoview 1000 Confocal microscope and an Olympus BX61 fluorescent microscope.

Label free quantitative proteomics of larval hemolymph
Shotgun quantitative proteomics was conducted on hemocyte-free hemolymph from larvae (n = 10) at 0, 6 and 24 h post infection with viable A. fumigatus (1 × 10 6 / larva). Protein (75 μg) was prepared, identified and analyzed according to established protocols and procedures [3].
Briefly, protein identification from the MS/MS data was performed using the Andromeda search engine in MaxQuant (version 1.2.2.5) and the data correlated against a 6-frame translation of the EST contigs for G. mellonella [56,57]. The MS proteomic data and MaxQuant search output files have been added to the ProteomeXchange Consortium [58] via the PRIDE partner repository with the dataset identifier PXD008196.
The Perseus software package (v. 1.5.5.3) was used for results processing, statistical analyses and graphics generation. LFQ intensities were log 2 -transformed and ANOVA of significance and t-tests between the hemolymph proteomes of 0, 6 and 24 h A. fumigatus treated larvae was performed using a p-value of 0.05 and significance was determined using FDR correction (Benjamini-Hochberg). Proteins which had non-existent values (suggestive of absence or very low abundance in a sample) were also used in statistical analysis Proteins found to be absent (below the level of detection) in one or more treatments and present (above the level of detection) in three or fewer treatments were also used in statistical analysis of the total differentially expressed group following imputation of the zero values with values that simulate low abundant proteins.

Statistical analysis
All experiments were performed on three independent occasions and results are expressed as the mean ± S.E. Analysis of changes in hemocyte density and protein abundance were performed by One-way ANOVA. All statistical analysis listed performed using GraphPad Prism. Differences were considered significant at p < 0.05.