Galleria mellonella - a novel infection model for the Mycobacterium tuberculosis complex

ABSTRACT Animal models have long been used in tuberculosis research to understand disease pathogenesis and to evaluate novel vaccine candidates and anti-mycobacterial drugs. However, all have limitations and there is no single animal model which mimics all the aspects of mycobacterial pathogenesis seen in humans. Importantly mice, the most commonly used model, do not normally form granulomas, the hallmark of tuberculosis infection. Thus there is an urgent need for the development of new alternative in vivo models. The insect larvae, Galleria mellonella has been increasingly used as a successful, simple, widely available and cost-effective model to study microbial infections. Here we report for the first time that G. mellonella can be used as an infection model for members of the Mycobacterium tuberculosis complex. We demonstrate a dose-response for G. mellonella survival infected with different inocula of bioluminescent Mycobacterium bovis BCG lux, and demonstrate suppression of mycobacterial luminesence over 14 days. Histopathology staining and transmission electron microscopy of infected G. mellonella phagocytic haemocytes show internalization and aggregation of M. bovis BCG lux in granuloma-like structures, and increasing accumulation of lipid bodies within M. bovis BCG lux over time, characteristic of latent tuberculosis infection. Our results demonstrate that G. mellonella can act as a surrogate host to study the pathogenesis of mycobacterial infection and shed light on host-mycobacteria interactions, including latent tuberculosis infection.


Background
Mycobacterium tuberculosis is a highly successful human pathogen that currently infects a third of the world's population, causing 9 million new cases of tuberculosis and 1.5 million deaths annually [1]. The close association of tuberculosis with the human immunodeficiency virus (HIV) pandemic, and the rising problem of multi-drug and extreme drug resistance have created difficulties for tuberculosis control programs worldwide [2,3]. There is thus an urgent need for a greater understanding of host and M. tuberculosis interactions, improved vaccines to prevent infection, and novel therapies to shorten the duration of treatment and target latent tuberculosis infection (LTBI), however this has in part been hampered by the current inadequate, costly and time-consuming in vivo infection models for tuberculosis. Animal models currently used in tuberculosis research include nonhuman primates (e.g. macaques), guinea pigs, rabbits, cattle, pigs, mice and zebrafish [4]. However working with such animals is difficult as the pathogenesis and progression of M. tuberculosis infection is complex and there is no single animal model that mimics all the aspects of pathogenesis in humans [5,6]. Mice are the most commonly used species (because of cost, availability of inbred lines, reproducibility of infection and availability of immunological reagents) [7], but granulomas and regions of hypoxia characteristic of LTBI, are not commonly seen in the lungs and no single mouse tuberculosis model can reflect the mechanistically and morphologically diverse forms of human tuberculosis [8]. The animal models most closely resembling human M. tuberculosis infection are non-human primates e.g. the macaque [9,10]. However ethical, cost, and practical considerations limit widespread use [6]. To overcome these issues there is an urgent need for a rapid, well-characterized, low-cost, high-throughput model that mimics the key features of M. tuberculosis pathogenesis to further understand hostpathogen interactions, that can also be used to screen novel drug candidates and assess the efficacy of vaccine candidates in early stage development.
Galleria mellonella insect larvae have increasingly been used as a surrogate to study host-pathogen interactions in a range of micro-organisms including Gram-positive and Gram-negative bacteria and fungi [11][12][13][14], and as a rapid model to screen novel antimicrobial drug candidates [15]. G. mellonella has a wide range of advantages that have made it a successful infection model. These include its sophisticated innate immune system, comprised of cellular and humoral defenses, that shares a high degree of structural and functional similarity to that of vertebrates [16,17], and its ability to discriminate between bacterial and fungal pathogens as identified by the repertoire of immune defense peptides [18]. G. mellonella are also easy to manipulate due to their large size (2-3 cm), allowing easy infection and collection of tissue/haemolymph for analysis. They are easy to maintain at 37°C making them well-suited to studying human pathogens, and precise infection studies can be performed by injection without anaesthesia for testing the efficacy of small quantities of antimicrobial agents. In addition, G. mellonella are ethically more acceptable than the use of vertebrates meaning larger group sizes can be used to enhance reproducibility [11,15,19,20].
Here we report the use of G. mellonella as a novel infection model for members of the M. tuberculosiscomplex. Due to safety and practical considerations when using M. tuberculosis as an experimental organism, we have carried out this study using the nonpathogenic, luminescent reporter vaccine strain, M. bovis BCG lux [21], which has been well characterized and validated in a wide range of in vitro and in vivo models and clinical studies (e.g. liquid broth, human whole blood and cells, mouse) and has the advantage of faster measurement of mycobacterial viability [22][23][24][25][26][27][28][29][30][31][32]. Over a time-course we have assessed the survival of G. mellonella larvae in response to infection with different inocula of M. bovis BCG lux, and monitored the survival of M. bovis BCG lux by measuring bioluminescence and correlating with CFU. Furthermore we have undertaken histopathology staining and transmission electron microscopy (TEM) analysis of M. bovis BCG lux infected G. mellonella to look at bacteria inside phagocytic haemocytes and study the pathogenesis and progression of mycobacterial infection. Our results demonstrate that this mycobacteria-G. mellonella infection model can be used to study mycobacterial pathogenesis, particularly in the context of granuloma formation, and has the potential to be used in the assessment of novel vaccine candidates and anti-mycobacterial agents. The availability of this novel mycobacterial infection model also has the potential to significantly reduce the number of experimental animals used in some areas of tuberculosis research.

Results
Survival analysis of G. mellonella following infection with M. bovis BCG lux G. mellonella survival (n = ≥ 10 per group), in response to differing inocula of M. bovis BCG lux infection (up to 2 × 10 7 CFU per larva) was determined over a 96 h period. Larvae that did not show any movement in response to touch were considered dead.
G. mellonella, infected with doses of 1 × 10 5 or 1 × 10 6 CFU M. bovis BCG lux, survived 100% (Figure 1 and  Supplementary Table 1) as did the 'pricked' (empty needle injection) and 'uninfected' (PBS-Tween 80) control  Figure 1. Effect of varying inoculum dose of M. bovis BCG lux on the survival of G. mellonella larvae following incubation at 37°C for 96 h. Healthy larva (n = ≥ 10 per group) infected with differing inocula of M, bovis BCG lux (up to 2 × 10 7 CFU per larva), were examined for survival by response to touch at 24 h intervals over a period of 96 h. Control groups (n = 10 larvae per group) included an uninfected group inoculated with 10 µl PBS-Tween 80, and a 'pricked' larval group (empty needle injection). Larvae were incubated at 37°C in the dark and monitored on a daily basis. Larvae that did not show any movement in response to touch were considered dead. Data are pooled from at least two independent experiments with a minimum of 10 larvae per group per run.
groups. This demonstrated that neither the needle injection nor the 0.05% Tween 80 used to prevent mycobacterial clumping, had any effect on G. mellonella survival. At higher doses of M. bovis BCG lux (1 x 10 7 and 2 × 10 7 CFU), a concentration-dependent effect on survival was observed. 50% G. mellonella survival was observed using a dose of 1 × 10 7 CFU M. bovis BCG lux per larva at 96 h, however there was less than 1% survival using a dose of 2 × 10 7 CFU per larva at this time-point. A time period of 96 h was used to measure larval survival, as at approximately 1 week of incubation at 37 C some of the larva in the 'pricked' and 'uninfected' groups started to pupate as they entered the next stage of the life cycle. Other observations of mycobacterial infected larvae compared to uninfected larvae over time, were a slight darkening in color and reduced flexibility and size, characteristic of the organism wasting.

M. bovis BCG lux survival in G. mellonella over time
The survival of M. bovis BCG lux (dose, 1 × 10 7 CFU) within G. mellonella was determined over a 2 week timecourse to further validate G. mellonella as an infection model for mycobacteria. Here we investigated the recovery of M. bovis BCG lux by luminescence measurement over time (0, 24, 48, 72, 96, 168 and 366 h) from homogenized larvae (n = 5 per time-point) to assess the ability of the immune system of G. mellonella to control mycobacterial growth. Only live, non-melanized larvae were used for homogenization.
When G. mellonella were inoculated with a dose of 1 × 10 7 CFU per larva there was approximately a 1 log reduction in M. bovis BCG lux luminescence between 0 and 168 h, with a further steady decrease in bacterial numbers until the end of the time-course at 366 h ( Figure 2). When we measured live G. mellonella uninfected crude extracts for background luminescence, very low levels were detected ranging from 5000-8000 RLU/ml indicating this did not have any impact on mycobacterial luminescence.
In addition to the luminescence measurement of mycobacteria, we also plated homogenized samples (n = 4 larva per time-point) for CFU quantification at the corresponding time-points up to 168 h on to Middlebrook 7H11 agar, containing hygromycin as an antibiotic selection marker and piperacillin which inhibited the growth of contaminating G. mellonella bacterial flora. Contamination from the normal bacterial flora of G. mellonella has been observed in other studies [33,34] but a recent paper by Entwistle & Coote [35] demonstrated that the antibiotic piperacillin was able to inhibit G. mellonella contaminating bacterial flora but did not inhibit the growth of non-tuberculous mycobacteria. We undertook a minimum inhibitory concentration (MIC) assay to assess the inhibitory concentration of piperacillin against M. bovis BCG lux at doubling concentrations ranging from 1280 -10 µg/ml. The MIC of piperacillin against M. bovis BCG lux was 160 µg/ml. A concentration of 10 µg/ml was found to inhibit the growth of contaminating G. mellonella bacterial flora and hence was added to agar plates for M. bovis BCG lux CFU quantification. We found that the RLU:CFU ratio was consistent over the time-course with an average ratio of 4:1 (range 2:1 to 5:1) (data not shown).

Histopathological analysis of M. bovis BCG lux infection of G. mellonella larvae
Granulomas in the human host, are highly dynamic and characteristic cellular structures that are the hallmark of tuberculosis and not only contain and control infection but also provide a niche for mycobacteria to survive. However, not all animal models form such structures in response to mycobacterial infection.
G. mellonella has an innate immune system comprised of cellular and humoral defenses including phagocytosis of pathogens by haemocytes (a similar cell type to human macrophages and neutrophils) [36] and the production of antimicrobial peptides, complement-like proteins and reactive oxygen and nitrogen species for killing ingested pathogens, all found within the haemolymph of G. mellonella [16]. Here we undertook both histopathology and TEM to follow the progression of M. bovis BCG lux infection, assess the impact of infection on larval tissues and determine whether granulomas are induced in this model of infection.
Histopathology analysis of sectioned larvae infected with M. bovis BCG lux, and stained using both H&E and ZN (acid-fast), demonstrated that at 24 h post-challenge, rod-shaped acid-fast bacilli (stained pink) were visible and many were co-localized with host cells (stained blue) ( Figure 3(A)), presumably haemocytes. We also observed isolation and compartmentalization of bacilli within granuloma-like structures ( Figure 3(B-D)). At 48 h postchallenge acid-fast amorphous material began to appear, most likely representing the by-product of destroyed acid-fast bacteria (Figure 3(D)). At subsequent time-points the acid-fast amorphous material within the granuloma-like structures increased, whilst individual bacilli were observed less frequently. By 168 h, the amorphous acid-fast material was the main structure associated with the granuloma-like structures, although occasionally a small number of individual bacilli were also observed ( Figure 3(E)). Staining serial sections with H&E demonstrated that the bacilli in the granuloma-like structures were colocalized with host cells (Figure 3(F,G)). In the center of the granuloma-like structures necrotic host cells were evident (Figure 3(F-H)) and in some cases brown pigmented material which co-localized with the amorphous acid-fast material was detected ( Figure 3(H)). Although using histology it was not possible to demonstrate that the bacilli were internalized, haemolymph smears stained with ZN appeared to demonstrate intracellular location of M. bovis BCG lux bacilli in individual host cells (Figure 3(I)) and in cell aggregates (Figure 3(J)), which have the appearance of granuloma-like structures. These structures were located in the haemocoel presumably surrounded by haemolymph, which was not possible to demonstrate in histological sections. Sometimes the granulomas were in contact with the fat body and infrequently with other internal organs ( Figure 3(K, L)). No ZN positive bacilli were identified in uninfected (pricked) larvae or in larvae challenged with PBS (Supplementary Figure 1).

Phagocytosis of M. bovis BCG lux infected haemocytes in vivo
To assess phagocytosis of M. bovis BCG lux (dose 1 ×

Discussion
The use of invertebrates, such as Caenorhabditis elegans, Drosophila melanogaster and G. mellonella, as models for infection studies have been widely reported as they are considerably cheaper, ethically less contentious than mammals, and although insects lack an acquired immune system they do possess a complex innate immune system [37]. Furthermore D. melanogaster has been used as an infection model for M. abscessus and M. marinum, particularly to study the pathogenesis and metabolic consequences of mycobacterial infection [38][39][40][41]. The larvae of G. mellonella have been widely used to study host-pathogen interactions for a range of microbes, including fungi, bacteria and viruses [11][12][13][42][43][44][45], but little has been reported using mycobacteria [35,[46][47][48], for which new in vivo infection models are greatly needed. Here we have demonstrated that this simple invertebrate model shows promise as a novel infection model for the M. tuberculosis complex.
Over 1000 articles have been published on the use of G. mellonella as a surrogate host infection model for studying the pathogenesis of micro-organisms, with more than 200 publications in 2014-2015 alone [15]. Advantages of using this insect larva include, no ethical implications, wide availability of larvae from commercial suppliers, low cost, simple housing and easy maintenance at 37°C, which also mimics the human host, large size (20-30 mm) allowing easy manipulation without the need for anaesthesia, precision injection into the haemocel through the last pro-leg, and the availability of host tissue/haemolymph for experimental analysis [11,15,19,20]. In addition, a good correlation has been established between the pathogenicity of several micro-organisms in G. mellonella with vertebrate models of infection [49]. Moreover, G. mellonella has an innate immune system, comprised of cellular and humoral defenses, that share a high degree of structural and functional similarity to that of vertebrates [16], and has been shown to discriminate between bacterial and fungal pathogens as judged by the repertoire of immune defense peptides [18]. A transcriptomic study of G. mellonella larvae found a large repertoire of immune effectors, many shared with mammals [50]. Hence bacteria can elicit an immune response in G. mellonella that is species-specific and is attractive for studying the pathogenesis of micro-organisms. This model also has the benefits of allowing infection experiments to be undertaken in a controlled manner with precise time-course measurements, analysis of changes in host-pathogen transcriptome, proteome and biochemistry, and far greater group sizes and numbers of replicates (n = 20 minimum per group) compared to animal studies (e.g. mouse n = 5 per group), enhancing reproducibility.
In this study we set out to evaluate the use of G. mellonella as an infection model for M. bovis BCG lux, a luminescent derivative of the vaccine strain that has the advantages of faster measurement of mycobacterial viability (luminescence requires co-factors only found in living metabolically active cells) compared to the slow growing M. tuberculosis with a doubling time of approximately 24 h [21,22,24]. Additionally, M. bovis BCG lux requires containment level (CL) 2, rather than CL3 laboratories. This organism has been very wellcharacterized and utilized in a wide range of clinical and non-clinical studies [21,24,[27][28][29][30][31][32]51]. For example, differences in M. bovis BCG lux RLU measurements have been shown to reflect changes in the immune response in human whole blood and cellular models, including differences between tuberculin-positive and -negative individuals [23], changes induced by HIV infection and following antiretroviral treatment [25,26], changes in mycobacterial growth limitation induced by vaccines [24] and Vitamin D supplementation in vitro and in a clinical trial [29][30][31], and differences in gene expression in immune biological pathways compared to M. tuberculosis [32]. In addition we have also evaluated luminescent mycobacteria in the mouse model [27]. Hence the use of M. bovis BCG lux in in vivo studies as a model organism for M. tuberculosis lends further benefit to in vivo assays using G. mellonella.
To characterize the infection model we first assessed the virulence of M. bovis BCG lux in G. mellonella using a range of doses, and determined survival of both host and mycobacteria over a 96 h time-course of infection. All G. mellonella survived at infection doses of 1 × 10 5 and 1 × 10 6 CFU of M. bovis BCG lux, whereas at higher doses a concentration-dependent effect was observed. A dose of 1 × 10 7 CFU was therefore selected for further time-kill, histopathology and TEM studies and would be suitable for testing drug treatments using this model.
The observation of 'wasting' in mycobacteria infected larvae, compared to uninfected, is consistent with the clinical feature of tuberculosis in humans and animals and warrants further investigation. This has also been described by Dionne and Schnieder in the D. melanogaster model of M. marinum infection [39].
Due to differences between bacterial species, strains, growth rates, pathogenicity and inocula used to infect G. mellonella and also time-points for larval survival measurements, comparing larval survival between bacterial species is difficult with, ideally, side by side experiments under the same conditions being carried out. However, at the 96 h time point with an equivalent dose of 1 × 10 7 CFU of M. bovis BCG lux or M. fortuitum NCTC 10394 [35], 50% and no larvae respectively, survived. Comparative survival may reflect differences in the mechanism(s) of pathogenicity or that M. fortuitum NCTC 10394 is a faster growing species.
We also assessed survival of M. bovis BCG lux over time-course of 366 h using a dose of 1 × 10 7 CFU per larva and observed approximately a 1 log reduction in luminescence between time-points 0 and 168 h postinfection which continued to steadily decrease until the end of the time-course, indicating persistent viable mycobacterial infection. Whilst G. mellonella larvae do not possess an adaptive immune system they do display an innate immune system similar to humans with many analogous functions including, phagocytosis of pathogens by haemocytes [17,34,[52][53][54], a similar cell type to human macrophages and neutrophils [36]. They also produce antimicrobial peptides, complement-like proteins and reactive oxygen and nitrogen species for killing ingested pathogens, all found within the haemolymph, a biological fluid similar to mammalian blood [15,16,18]. We speculate that death of infected larvae with higher M. bovis BCG lux doses (10 7 and 10 6 CFU per larvae) occurs despite no increase in mycobacterial growth, as a result of an overwhelming G. mellonella immune response at the site of infection. Further investigation of the molecular and functional immune mechanisms and specific cells involved in suppression of mycobacterial luminescence will be the subject of future work.
Luminescence measurement has been shown to be a measure of mycobacterial metabolic activity, related to viability [21,22,24,33]. We also isolated and enumerated mycobacterial CFU from infected G. mellonella at the same respective time-points as luminescence measurements up to 168 h to determine the correlation between RLU and CFU in the G. mellonella model. G. mellonella homogenized larvae were plated on to Middlebrook 7H11 agar, specific for mycobacterial growth, incorporating hygromycin as an antibiotic selection marker for the lux reporter and piperacillin to inhibit the growth of contaminating bacterial flora from G. mellonella. We found the MIC of piperacillin against M. bovis BCG lux to be 160 µg/ml and whilst inhibiting the growth of G. melonella bacterial flora, it did not inhibit the growth of M. bovis BCG lux at the concentration used (10 µg/ml). We found that the RLU: CFU ratio for M. bovis BCG lux correlated well at each time-point over the timecourse with an average RLU: CFU ratio of 4:1 which is also consistent with the RLU: CFU 3:1 ratio in liquid broth culture. These findings highlight that although luminescence is a measure of metabolic activity in M. bovis BCG lux, it correlates with CFU in vivo in the G. mellonella model over time and further highlights the benefit of using this luminescent reporter organism.
Similarly the experiments of Meir et al. [48] showed that there was correlation between lux RLU and CFU with the fast growing non-tuberculous M. abscessus in the background of G. mellonella, as has also been shown with other bacteria e.g. Enterococcus faecalis [55].
The hallmark of tuberculosis infection in humans is the granuloma; a highly organized and dynamic cellular structure where mycobacteria reside in a non-replicating dormant state known as LTBI [6]. Several types can be found in humans including the caseous, non-necrotizing, and fibrotic granulomas [56]. However studying mechanisms underlying latency and reactivation of tuberculosis is hampered by the limitations of current animal models [57]. Furthermore the design of new drugs which penetrate granulomas to target LTBI in particular, is an urgent priority for any treatment control program [58]. Recently the zebrafish tuberculosis model, using M. marinum, has been used to mimic aspects of LTBI due to its small size, rapid reproduction, formation of granulomas and advanced genetic tools, making it a suitable animal model for large-scale screening of novel therapeutic agents in early-stage preclinical studies. However, zebrafish are anatomically and physiologically different to humans and lack the clinical manifestations and symptoms of tuberculosis disease and moreover they cannot be infected with the pathogenic human strain M. tuberculosis [4,[59][60][61][62].
Granuloma-like structure formation in G. mellonella has been shown for the pathogenic fungal infections Paracoccidioides lutzii and Histoplasma capsulatum [34]. Hence, we undertook both histopathology staining and TEM to visually assess the progression of infection and examine the changes in host tissue and induction of granuloma-like formations in response to infection. Our data show that M. bovis BCG lux challenged larvae induce granuloma-like structures to sequester the bacteria as early as 24 h following infection, and inside these formations the bacilli lose their characteristic morphology forming amorphous acid-fast bodies by 48 h. Analysis of infected phagocytic haemocytes, extracted from G. mellonella larvae, show internalization and aggregation of M. bovis BCG lux in granuloma-like structures, with increasing accumulation of lipid bodies within M. bovis BCG lux over time. The accumulation of lipid bodies, in the form of triacylglycerides (TAGs), is evident in M. tuberculosis residing within lipid-rich foamy macrophages in human granulomas [63,64], and is proposed to serve as a source of carbon and energy representing a hallmark of persistent and non-dividing mycobacteria, labelled the drug tolerant, 'fat and lazy' bacilli [63,65]. It is therefore no surprise using this infection model, that mycobacteria are highly likely sourcing lipid rich material within their granuloma-like structures from their fatty G. mellonella host. These granuloma-like structures are formed in response to challenge with M. bovis BCG lux, which is an attenuated M. bovis strain. It would be of significant interest to study the response of G. melonella larvae to pathogenic mycobacterium (e.g. M. tuberculosis, M. bovis) as the vast majority of data in mammals derives from exposure to these species.
Whilst further investigation is required into the nature and composition of both the granuloma-like structures formed in response to M. bovis BCG lux infection, and the intracellular mycobacterial lipid bodies, our results are highly encouraging that G. mellonella represent an organism capable of forming such pathological features of tuberculosis, which are highly desirable for a novel tuberculosis infection model. In addition a range of mycobacterial strains, mutants classes (e.g. lipids, metabolism, persistence, stress) and reporters (e.g. luminesence and fluorescence), to assess host-pathogen interactions and characterize mechanisms of pathogenicity, could be used in this model.
In summary, this is, to our knowledge, the first report of G. mellonella larvae being used as a suitable infection model for a member of the M. tuberculosiscomplex. Importantly we show that G. mellonella replicates features of the pathogenesis of tuberculosis through the induction of granuloma-like structures characteristic of LTBI, and the inclusion of lipid bodies within mycobacteria over time, characteristic of persistent infection. Further evaluation of this model has the potential to markedly reduce the use of more expensive and time-consuming animal models for evaluating mycobacterial host pathogenicity, the toxicity and efficacy of novel anti-mycobacterial agents and novel vaccine candidates in vivo [66]. The characterization and optimization of this model with both non-pathogenic, and pathogenic drug-sensitive and drug-resistant M. tuberculosis isolates will form the basis of future studies.

Optimization and preparation of M. bovis BCG lux inoculum
Prior to infection experiments, a vial of mycobacteria was defrosted and grown in liquid medium (as described above) to mid-log phase for 72 h with M. bovis BCG lux growth monitored by luminescent measurements on a daily basis [22]. Cultures were centrifuged at 3880 g for 10 minutes, supernatant decanted and the cell pellet was resuspended in phosphate buffered saline (PBS) (Sigma-Aldrich) containing 0.05% (v \v) Tween 80 (PBS-Tween 80) and washed a further twice. PBS-Tween 80 was found to prevent mycobacterial clumping particularly when achieving dense concentrations of mycobacterial inoculum (e.g. 10 9 RLU/ ml). The washed cell pellet was diluted in PBS-Tween 80 and the required mycobacterial inoculum measured by luminescence, resulting in a homogenous suspension. The inoculum dose was verified by plating on to Middlebrook 7H11 agar.

In vivo virulence assays of G. mellonella infected with M. bovis BCG lux
Final instar larvae of G. mellonella were obtained from Livefoods Direct Ltd, Sheffield, UK and stored in wood shavings in the dark at 19°C. Healthy larvae were identified by their cream colour, without dark discolouration, the ability to right themselves if turned on their backs, a body weight of approximately 250 mg and body length of 2-3 cm, and were used in infection experiments within 3 days of delivery. For experimental use healthy larvae were counted into 25 cm Petri dishes containing a layer of filter paper. Once selected, no food was made available. Larvae displaying signs of death (no movement when touched) were discarded along with those undergoing pupation.
Prior to each experiment, larvae were surface disinfected with ethanol, and at least 10 larvae per group were injected with 10 μl of a M. bovis BCG lux suspension in PBS-Tween 80. Doses, between 1 × 10 5 and 2 × 10 7 CFU, were injected via the last left pro-leg into the haemocoel using a SGE 25 µl microvolume syringe. Control groups (n = 10 larvae) included an uninfected group inoculated with 10 µl PBS-Tween 80, and a 'pricked' larval group (empty needle injection). Larvae were maintained in petri dishes containing a base layer of filter paper at 37°C in the dark. All larvae were monitored on a daily basis and melanization and survival were recorded. Larvae that did not show any movement in response to touch were considered dead. Data are pooled from at least two independent experiments with a minimum of 10 larvae per run. Kaplan-Meier survival curves of G. mellonella in response to pricked, uninfected and M. bovis BCG lux infection were plotted to display percentage survival over time.

Recovery of M. bovis BCG lux from G. mellonella over time
At least 30 larvae were infected with a 10 µl dose of 1 × 10 7 CFU as described above. At 24, 48, 72, 96, 168 and 366 h, five live larvae per time-point (no melanization) were disinfected by wiping down with 70% ethanol and homogenized in PBS-Tween 80 using a FastPrep-24 5G (MP Biomedicals) with sterile metal beads. Mycobacterial luminescence (RLU/ml) was measured in the homogenized lysate as described above. The mean RLU/ml of M. bovis BCG lux survival in G. mellonella was determined over 366 h. In addition, CFU enumeration was also determined in homogenized lysate by plating serial dilutions in duplicate on to Middlebrook 7H11 agar plates containing 50 µg/ml hygromycin as an antibiotic selection marker and 10 µg/ml piperacillin (Sigma) to inhibit the growth of G. mellonella bacterial flora. The mean CFU/ml of M. bovis BCG lux survival in G. mellonella was determined at each time-point over 168 h.
To determine CFU enumeration of the inoculating M. bovis BCG lux, 100 µl of mycobacterial inoculum was plated on to Middlebrook 7H11 agar plates, containing 50 µg hygromycin.

Minimum inhibitory concentration (MIC) assay of piperacillin
The MIC of piperacillin was determined against M. bovis BCG lux. Two-fold serial dilutions of piperacillin ranging from 1280 to 10 μg/ml were prepared in Middlebrook 7H9 broth and 100 μl of each concentration was combined with an equal volume of log-phase mycobacterial culture diluted to~10 6 CFU/ml in a 96 well plate. Growth media containing mycobacteria only was used as the positive control and growth media alone as the negative control. For each concentration and control was performed in triplicate. Plates were incubated in a rocking incubator at 20 rpm at 37̊C for 7 days. The MIC was defined as the lowest concentration at which no mycobacterial growth was observed visually or spectrophotometrically using a plate reader (Molecular Devices, UK) at an optical density 595 nm. Each MIC experiment was performed in triplicate.
To determine the concentration of piperacillin required to inhibit the growth of G. mellonella bacterial flora, 50 µl homogenized larvae (n = 5 larva) was plated on to Middlebrook 7H11 agar plates containing a range of concentrations of piperacillin (10, 20, 40 µg/ml). Agar plates without piperacillin were used as the positive control. Plates were incubated at 37°C for up to 2 weeks to check for growth of G. mellonella bacterial flora.
Time course, histopathology and transmission electron microscopy (TEM) analysis of M. bovis BCG lux recovered from G. mellonella The time-course study was undertaken to investigate the progress of M. bovis BCG lux infection (dose 1 × 10 7 CFU) within G. mellonella infected larvae, and assess the impact of infection on G. mellonella internal tissue. Three larvae per time-point per group were used for histopathology analysis and at least ten larvae per timepoint for TEM. The time-course ran for 96 h postinfection, with time-points at 0 h and then every 24 h intervals. In addition, some infected larvae were sacrificed and processed at 144 h and 336 h after inoculation.
For histopathological analysis, infected and uninfected (control group) larvae were fixed by injecting 100 µl of buffered formalin with an insulin syringe into the last left proleg. Larvae were kept in 40 ml formalin at room temperature. Subsequently, each larva was cut into two parts along the midline. Both parts were kept and processed for histology (Sakura Tissue-Tek VIP), embedded in paraffin wax (Thermo Scientific HistoStar embedding center), sectioned at 4 µm using a Leica RM2135 microtome, and then stained with haematoxylin and eosin (H&E), and Ziehl-Neelsen (ZN) stain, a specific stain for acid fast bacteria. The tissue slides were examined under a light microscope (Nikon Eclipse 80i) and photographed using a digital camera (Nikon DS-Ri1) attached to a computer equipped with NIS-Elements BR imaging software.
For TEM, at least 10 larvae per time-point (0, 24, 48, 72, 96, 144, 366 h) per group were used for haemolymph extraction. Haemolymph was collected from the haemocoel by puncturing the larvae in the last left pro-leg using a sterile needle. Haemolymph was carefully squeezed out of the larvae by applying gentle pressure, pooled into 50 ml of 1% formaldehyde and kept at room temperature for 2 h. Cells were then centrifuged and the cell pellet was re-suspended in 4% glutaraldehyde buffer and stored at 2-8C. An aliquot of the cells was then placed onto positively charged slides for ZN staining. The remaining cells were post-fixed with 1% osmium tetroxide (OsO4). After dehydration and embedding of the samples, ultra-thin (70-90 nm) sections were mounted on to grids and stained with uranyl acetate (0.5% w/v) and lead citrate (3% w/v) prior to examination using a Tecnai bioTWIN transmission electron microscope.