Using the wax moth larva Galleria mellonella infection model to detect emerging bacterial pathogens

Climate change, changing farming practices, social and demographic changes and rising levels of antibiotic resistance are likely to lead to future increases in opportunistic bacterial infections that are more difficult to treat. Uncovering the prevalence and identity of pathogenic bacteria in the environment is key to assessing transmission risks. We describe the first use of the Wax moth larva Galleria mellonella, a well-established model for the mammalian innate immune system, to selectively enrich and characterize pathogens from coastal environments in the South West of the UK. Whole-genome sequencing of highly virulent isolates revealed amongst others a Proteus mirabilis strain carrying the Salmonella SGI1 genomic island not reported from the UK before and the recently described species Vibrio injenensis hitherto only reported from human patients in Korea. Our novel method has the power to detect bacterial pathogens in the environment that potentially pose a serious risk to public health.


Introduction
Emerging infectious diseases (EIDs) pose a major threat to human health 1 . A large proportion of EIDs are caused by bacteria (estimated to be 54% 1 and 38% 2 ). Although most emerging bacterial pathogens have zoonotic origins, a large proportion of infectious bacteria are free-living, for instance being associated with food 3 , drinking water 4 or recreational waters 5 . Microbial safety is routinely assessed through the quantification of Faecal Indicator Bacteria (FIB) 6 . However, many FIB lineages are not associated with disease and there is no 3 a priori reason to expect a relationship between FIB abundance and non-gastrointestinal disease (e.g. ear or skin infections). There are dozens of bacterial genera occurring in natural environments that are not primarily associated with human or animal faecal contamination but that are able to cause opportunistic infections (e.g. 7 ). Alternatives to FIB such as quantification of pathogen-specific genes via molecular methods 8 , flow cytometry (e.g. 9 ) or isolation of specific pathogens (e.g. 10 ) either are not linked to infection risk, are based on costly methodologies or are limited to a subset of 'known knowns'. The current lack of a direct screening method for the presence of pathogenic bacteria in environmental samples is therefore a major barrier to understanding drivers of virulence and ultimately infection risk.
We demonstrate the use of the Wax moth larva Galleria mellonella as a bioindicator for microbial water quality, and a means to selectively isolate and characterize pathogens. G.
mellonella is a well-established model system for the mammalian innate immune system and has been used extensively to test for virulence in a range of human pathogens by quantifying survival rate after injection of a defined titre of a specific strain or mutant 11 . Bacterial virulence in Galleria is positively correlated with virulence in mice 12 as well as macrophages 13 . Instead of quantifying the virulence of a specific bacterial clone, here we measure Galleria survival after injection with entire microbial communities from concentrated environmental water and sediment-wash samples. We isolate bacteria responsible for Galleria mortality and assess their pathogenic potential through whole-genome sequencing.

Results
Our survival assay shows that Galleria mortality after 72 hours varied widely between both water and sediment samples collected at two dates from eight locations across Cornwall (U.K.), ranging from 5% to 95% (Fig. S2) We chose four environmental samples exhibiting high (≥ 70%) Galleria mortality to isolate the clone(s) responsible for infection and reinoculated these samples to isolate bacteria from the haemocoel of infected, freshly killed larvae. All samples yielded a single colony type on each agar type, indicating that infections were (largely) clonal. A single clone was picked for each sample, grown up and assayed using three inoculation densities (1x10 2 CFU, 1x10 4 CFU, and 1x10 6 CFU) ( Fig. 1). All clones displayed high levels of virulence and were characterized using whole-genome sequencing (Fig. 1). We specifically focused on the identification of virulence-and antibiotic resistance genes (ARGs) as compiled in the VFDB 14 and CARD 15 databases respectively.
The first clone, isolated from estuarine mud (Supplemental Results) was identified as the enteric species Proteus mirabilis, most closely related to pathogenic strain HI4320 16 (Fig.   1A). Interestingly, this strain was found to carry a multidrug resistance genomic island (SGI1), first identified in an epidemic Salmonella enterica serovar Typhimurium clone in the 1990s 17 . This island has since been found in P. mirabilis isolated from human patients as well as from animals 18 but to our knowledge not from Proteus strains isolated from natural environments. No virulence genes were found using a 90% similarity cut-off, but several were identified using a 75% cut-off (Table S2). The clone contains several antibiotic resistance genes (ARGs), including the tetracycline efflux protein TetJ and AAC(6')-Ib7, a plasmid-encoded aminoglycoside acetyltransferase (90% similarity cut off, Table S3).
The second clone, isolated from beach sand, was found to belong to Vibrio injenensis, a recently described species only known from two strains isolated from human patients in  (Fig. 1B). The UK clone was 99% similar to the type strain M12-1144 T and carried 441 genes not present in the Korean strain. Both strains carry the rtx toxin operon (Table S4).
Only two ARGs, including tetracycline resistance tet34, could be identified at a 75% similarity cut off in the UK isolate (Table S5) aminoglycoside efflux pumps, a class C ampC beta-lactamase conferring resistance to cephalosporins and pmrE implicated in polymyxin resistance (Table S9).

Discussion
Our study utilized the low-cost and ethically expedient Galleria infection model to directly measure the presence of pathogenic bacteria in environmental samples without any prior knowledge of identity. As expected, some samples with low FIB counts contained pathogenic bacteria and some samples with high FIB counts showed low Galleria mortality ( Fig. S4). We note that of four pathogenic isolates, only one was a coliform and only two were gut-associated bacteria. Two out of the four isolates have not been reported from the U.K. before and potentially represent EIDs. This highlights the fact that transmission risk extends beyond 'usual suspects' and includes environmental-and largely uncharacterized strains. Our relatively simple methods can provide a basis for future studies to detect pathogenic bacteria in diverse environments, to ultimately elucidate their ecological drivers and estimate human infection risk.

Sample collection and processing
Eight sampling sites on the Fal estuary and English Channel coast near Falmouth (Cornwall, U.K.) were selected for water and sediment sampling. Four sites were estuarine and four truly coastal ( Figure S1). Samples from each site were collected on 21 June 2017 and 6 July 2017.
Each sampling effort consisted of collecting water (2 x 50 mL) and upper 1cm layer of sediment (~25 g) from all eight sites within a time span of two hours around low-tide.
Samples were collected using sterile 50mL centrifuge tubes. All samples were kept on ice during transport and processed in laboratory within an hour from collection. Larvae were scored as dead if they did not respond to touch stimuli by blunt sterile forceps.

Galleria mellonella assay for individual clones
To isolate and identify responsible pathogens, a subset of samples showing high Galleria mortality were inoculated and incubated again as described above. Eight live larvae showing signs of infection (melanisation) were selected at random from each sample group and dissected for haemocoel collection. Before haemocoel collection, the Galleria larvae  Contigs were ordered by whole genome alignment against a reference genome using progressive mauve (Darling et al 2010). The E. coli isolate was aligned against K-12 substr.
MG1655 (NC_000913.3), the P. mirabilis isolate was aligned against HI4320 (NC_010554.1), and the P. aeruginosa against PAO1 (NC_002516.2). The V. injenensis genome could not be aligned to a closed reference genome and so the contigs could not be ordered or separated by chromosome.

Data Availability
Trimmed reads and assemblies have been uploaded at the NCBI, Accession: PRJNA473311.
Raw data will be deposited into an appropriate database upon acceptance.