B. pseudomallei strain R15 (Lee et al., 2007) was used to construct the transposon mutant pool. B. pseudomallei wild type and mutant strains were routinely grown in Luria Bertani (LB) broth or Brain Heart Infusion (BHI) broth at 37°C with shaking at 250rpm. Tetracycline (100 μg/mL) and gentamycin (4 μg/mL) were used for selection when necessary.

The C. elegans strain rrf3(pk1426);glp-4(bn2) was propagated and maintained at 16°C on nematode growth medium (NGM) agar plates with Escherichia coli OP50 as the standard food source (Wong et al., 2018). The C. elegans population was treated with sodium hydroxide and alkaline hypochlorite to release embryos and all for age synchronization of the worms. The nematodes were propagated at 25°C until the young adult stage before they were used for infection with B. pseudomallei.

We used the EZ-Tn5<TET-1> Insertion Kit (Cat. No. EZ1921T, Epicentre, Illumina) to construct the transposon mutants, Transposome was prepared according to the manufacturer’s instructions and 1 μL of transposome was added to 50 μL of electrocompetent cells followed by incubation on ice for 30 minutes. The cell suspension was transferred to a pre-chilled electroporation cuvette (2-mm-wide gap) (Bio-Rad, Laboratories, California, USA) and pulsed using the Gene Pulser (Bio-Rad Laboratories, California, USA) at 2.5 kV, 25 μF and 200 Ω. Following the electroporation, SOC medium (950 μL) was added to the electroporated cells which were then incubated at 37°C with agitation for 2 hours. The transformed cells were spread on 2X LB agar supplemented with appropriate concentrations of tetracycline and further incubated at 37°C for 24 hours. The number of mutants obtained was estimated by counting the bacterial CFU. Once colonies had been washed with BHI broth and scraped off the plates, they were stored at -80°C in 30% glycerol. Multiple batches of electrotransformations were conducted until 500,000 mutants were obtained. A large-scale transposon mutant library was generated by combining these multiple pools of mutants (input pool). Aliquots of this input pool were stored at -80°C.

Approximately 1x10¹⁰ mutants from the input pool were inoculated into 50 mL fresh LB broth and grown at 37°C with aeration for 24 hours (T₁). Subsequently, 50 μL of this culture (T₁) was inoculated into 50 mL of fresh LB media and grown for a further 24 hours (T₂) followed by a second passage (T₃). The mutants were also grown for 24 hours at 37°C in M9 minimal medium (supplemented with 0.4% glucose as carbon source) (T₁) and then 1 mL of the culture was transferred to fresh medium for another 24 hours (T₂), followed by a third passage (T₃) for 24 hours (schematic diagram is shown in Supplementary Figure S1 ). For each output pool, genomic DNA was extracted from 5 mL of bacteria culture.

Approximately 1 x 10¹⁰ CFU transposon mutants (hereafter, referred to as the input pool) were inoculated into 50 mL LB broth and incubated at 37°C for 16 hours. One hundred microliters of the culture was spread evenly on NGM agar prepared in 6 cm Petri dishes. An additional 5 mL of the culture was reserved as output-LB. The NGM plates spread with the input pool were incubated at 37°C for 24 hours after which they were left to equilibrate to room temperature for a further 24 hours. The bacterial lawn was washed from 1-2 plates and collected as output-NGM. We also considered the possibility of potential changes to the in vitro grown population introduced during the bacterial preparation steps which may have led to the loss of some mutants from the input pool that was later subjected to in vivo selection. Mutants grown in both LB and NGM were harvested to gauge possible changes to the in vitro grown population. To infect C. elegans with the mutants, age-matched young adult worms (rrf3(pk1426);glp-4(bn2)) were shifted onto the equilibrated input pool mutants that were spread on the NGM agar. Plates were incubated at 25°C over the infection period ( Supplementary Figure S1 ).

At 6 and 24 hpi, mutants were harvested from worm intestines (Ooi et al., 2012). At the designated time points, the worms were rinsed with M9 buffer and transferred into microcentrifuge tubes. The tubes were briefly centrifuged at low speed to remove excess buffer. Worms were treated with 25 mM Levamisole (Sigma-Aldrich, USA) to anesthetize them, rinsed three times in 200 μL of 25 mM Levamisole and 1000 μg/mL kanamycin. The worms were incubated for 2 hours to kill the bacterial cells bound to the worm cuticle. Then, worms were washed four times with 200 μL of Levamisole (25 mM) to remove the killed bacteria and residual antibiotics. Worms were homogenized by mechanical disruption using a motorized pestle in 100 μL of 1% Triton X (Sigma-Aldrich, USA). Serial dilutions were performed on the worm lysates and 10 μL of each dilution was spotted on Ashdown agar to estimate the number of intestinal mutants recovered from each tube. The remainder lysate was spread on modified LB agar (3% tryptone, 1% yeast extract and 1.5% bacteriological agar) supplemented with tetracycline and incubated for 24 hours at 37°C. Colonies were soaked with LB broth, scraped off and harvested as in vivo output pools (output-6 hpi and output-24 hpi) ( Supplementary Figure 1 ).

All input and output pools were used to prepare sequencing libraries using the method previously described (Wong et al., 2016; Wong et al., 2018). In brief, 1μg of genomic DNA was fragmented to 100-1500 bp fragments and the fragments were end repaired and A-tailed with the NEBNext DNA library adapter. The MP_Ad_a and MP_Ad_b oligos were annealed together to prepare the adapter for the A-tailing of the DNA fragments. The adapter-ligated DNA fragments (200 ng) were subjected to amplification over 22 cycles with the Tra_Fp and Tra_Mp_Rp primers which contain unique 7-bp indexes (indicated by N) for sample multiplexing (Meyer and Kircher, 2010). Size selected (200 bp-500 bp) amplified libraries were purified and quantified with the Bioanalyzer and by quantitative real time PCR using primers qPCR_P5 and qPCR_P7. The purified libraries were sequenced on the Illumina HiSeq 2000 platform using the index primer Tra_IndP and the conventional sequencing primer Tra_SeqP. All primer sequences and PCR cycling conditions are available in Supplementary Table 1 .

Sequence reads within the Illumina FASTQ files were filtered based on 10 bases (TAAGAGACAG) matching the 3’ end of the transposon with one mismatch allowed. Matched reads were then edited to remove the transposon tag and the edited reads were mapped to the B. pseudomallei R15 reference genome sequence (Ghazali et al., 2021; European Nucleotide Archive: accession number ERS3410623) using SMALT-0.7.2. The exact transposon insertion site was determined using Bio::Tradis (https://github.com/sanger-pathogens/Bio-Tradis). The method of Langridge et al. (2009) was used to assess gene essentiality. Log₂ likelihood ratios (LLR) were calculated between the essential and non-essential models. If a gene had a LLR of< -2, it was classified as essential; alternatively, if it had a LLR of > 2, the gene was classified as non-essential. The log₂ fold change ratios of the observed reads were calculated based on Langridge et al. (2009) and used to compare the differences in read abundance between the input pool and in vivo output pools. For every gene (g), the log₂ fold change was calculated as log₂[(n_g,A + 100)/(n_g,B + 100)], where n_g,A is the number of observed reads in the output pool and n_g,B is the number of reads for the input pool (Wong et al., 2016). For genes with very low numbers of observed reads, high scores were smoothed out with a correction of 100 reads. A normal mode was fitted to the log₂ fold change mode of distribution and according to the fit, P-values were calculated for each gene. Genes with log₂ fold change of< -2 and corrected P-value of< 0.05 were considered to be essential for bacteria to grow under a particular condition.

The genes which met the above criteria and proposed as essential genes were subjected to a search against the collection of bacterial essential genes available in the Database of Essential Genes (DEG) (http://origin.tubic.org/deg/public/index.php/; version 14.7, - Oct. 24, 2016) (Luo et al., 2014). The search was performed using BLASTP and the default parameters provided in DEG. From the BLASTP search, similarities at protein-protein level with E-values of 10⁻¹⁰ or less were considered matches. Known B. pseudomallei virulence factors were retrieved from the Virulence Factors Database (VFDB) (http://www.mgc.ac.cn/VFs/; database version Feb. 11, 2017) (Chen et al., 2005) and the Burkholderia Genome Database (http://www.Burkholderia.com) (Winsor et al., 2008).

The deletion mutant of the selected target gene bpsl3313 was generated using the method described by Wong et al. (2018). The successful deletion of the targeted region was verified with Sanger sequencing. All primer sequences are listed in Supplementary Table 1 .

For the second selected target gene, bpsl2988, the insertion mutant bpsl2988::Tn5 was obtained from a B. pseudomallei strain R15 transposon mutant library (not published) constructed using the EZ-Tn5<TET-1> Insertion Kit (Cat. No. EZ1921T, Epicentre, Illumina). The transposon Tn5 insertion within the bpsl2988 open reading frame was confirmed and validated via Sanger sequencing.

Wild type B. pseudomallei R15 and mutant strains were grown in LB broth for 16 hours at 37°C. The cultures were initially adjusted to an OD_595nm of 0.5 in LB or M9 minimal medium, after which, 500 μL was transferred into 50 mL fresh media (LB or M9 minimal) and incubated with shaking at 250 rpm at 37°C. Serial dilutions of the cultures were plated every 2 hours to enumerate the number of live bacteria. For each dilution, triplicate aliquots of 10 μL were dropped onto Ashdown agar and plates were incubated for 48 hours at 37°C.

The CFU assay was performed to determine if the B. pseudomallei mutants were able to colonize C. elegans compared to the wild type strain. Intestinal bacteria were collected from infected worms using a method described above for harvesting mutants from the infected nematode intestine (Section 2.5). In brief, B. pseudomallei wild type and mutant overnight cultures were adjusted to an optical density (OD_595nm) of ~ 1.5. Twenty μL of each standardized bacterial inoculum was spread evenly on NGM agar in 3.5 cm Petri dishes and incubated at 37°C. The age-matched rrf3(pk1426);glp-4(bn2) young adult worms were transferred onto individual wild type or mutant strain lawns on NGM agar. At various time points (6, 24, 48 and 72 hpi), 10 worms were picked from each plate and transferred into microcentrifuge tubes containing 100 μL of 25 mM Levamisole. For each bacterial strain, 3 tubes (technical replicates) with each consisting of 10 worms were prepared. Worms were washed, and a motorized pestle was used to homogenize the worms in 50 μL of 1% Triton X. Serial dilutions were performed on the lysates and each dilution was spotted on Ashdown agar in triplicate. Agar plates were incubated at 37°C and colonies were counted after 48 hours of incubation. The number of bacteria per worm was obtained by dividing the total number of bacterial CFU in 50 μL of worm lysate with the number of worms retained in each tube.

Overnight cultures of the B. pseudomallei strains were adjusted to OD_595nm = 1.5 and 20 μL of each bacterial suspension was spread evenly on NGM agar in 3.5 cm Petri dishes and incubated as previously described. E. coli OP50 was used as a control for the assay. Forty age-matched young adult worms were transferred to the bacterial lawn (triplicate plates for each test strain) and plates were further incubated at 25°C. The number of live and dead worms were scored every 4−6 h. Nematodes were assumed to be dead when they failed to respond to touch (Huber et al., 2004).

Colonies of wild type and mutant strains were inoculated into BHI broth and incubated at 37°C. Overnight cultures were washed twice with PBS, adjusted to an OD_595nm of 0.5 (approximately 1 x 10⁸ CFU/mL) and then diluted 1:200 in sterile Phosphate Buffered Saline (PBS). Female BALB/c mice, aged 8 – 10 weeks old, were sourced from the Animal House Facility, Universiti Kebangsaan Malaysia. Mice were maintained under specific-pathogen-free conditions in a positive pressure environment at 20 – 25°C, subjected to a 12-hour light/dark cycle and fed with a protein-enriched diet and water ad libitum. Mice were divided into groups of five and for each group, mice were injected intraperitoneally with either wild type or different mutant strains at 1 x10⁵ CFU of bacteria (in 200 μL of PBS). All infected mice were euthanized at 24 hpi by ether inhalation. The lungs, livers and spleens were aseptically removed and individually homogenized in 2 mL of sterile ice cold PBS. Organ homogenates were serially diluted and for each dilution, 10 μL was spotted on Ashdown agar (in triplicate). Colonies grown on plates were counted after 2 days of incubation at 37°C and the bacterial load per organ was determined. The animal experiments were approved by the Universiti Kebangsaan Malaysia Animal Ethics Committee (UKMAEC; FST/2016/SHEILA/23-MAR/732-MAR-2016-OCT-2018) and strictly adhered to the Universiti Kebangsaan Malaysia Animal Ethics Guidelines.

The number of colony forming units (CFU) were analysed with GraphPad Prism version 5.04 and the Mann-Whitney U-test was used to determine statistical significance. The doubling time (g) based on growth curves, was calculated from the exponential phase using the formula:

where N₀ = number of CFU at a point during log phase, N_t = number of CFU at a different time point during log phase and t = time interval between N₀ and N_t. The in vitro growth analysis data were expressed as mean ± standard error of the mean (SEM) from at least two independent assays. Statistical analyses were performed using the unpaired, two-tailed Student’s t-test. Results obtained from worm killing assays were analyzed using the Kaplan-Meier nonparametric survival analysis in Statview^® 5.0.1 (SAS Institute, Inc). All data are presented as mean ± standard deviation (SD) of a representative from at least three independent experiments.

Sequence reads were deposited in the European Nucleotide Archive database with accession number PRJEB13678 and are accessible via http://www.ebi.ac.uk/ena/data/view/PRJEB13678. The sample accession numbers are ERS1124824(Input A), ERS1124825 (Input B), ERS1124812 - ERS1124814 (LB_T1 – LB_T3), ERS1124818 - ERS1124820 (M9_T1 – M9_T3), ERS1124808 (LB), ERS1124809 (NGM), ERS1124810 (6 hpi) and ERS1124811 (24 hpi).

A large-scale mutant library was created by electroporating transposome (Tn5 transposon-transposase complex) into B. pseudomallei and this process generated multiple batches of mutants which were pooled to achieve a large-scale library of ~500,000 mutants. This library is henceforth referred to as “input pool” to represent the initial library. A custom sequencing primer was designed to amplify the last 10 bp sequence of the transposon extending into the adjacent bacterial genomic sequence to precisely identify transposon insertion sites (TIS) for each mutant (Langridge et al., 2009). An average of 20 million reads were obtained for each sequencing library, and the 10 bp transposon tag was detected in more than 92% of the total reads. All sequencing reads without the transposon tag were removed before the sequences were mapped. About 78% of the transposon-tagged reads from replicate libraries (Input A and Input B) were uniquely mapped to the B. pseudomallei R15 genome. The number of unique TIS identified across the genome was about 413,000 for Input A and 737,000 for Input B ( Table 1 ). Correlation analysis revealed a Spearman’s rho correlation coefficient of 0.9855, indicating low variability between the two replicates (Input A and Input B) ( Supplementary Figure S2 ). We combined both Input A and Input B reads and this increased the sequencing depth which enhanced the degree of confidence in identifying unique TIS. A total of 848,811 unique TIS were identified which were distributed across the B. pseudomallei R15 genome. The highest density of insertions (about 618,812 unique TIS) was detected in chromosome 1 at an average of one TIS every 7 bp. Approximately 229,999 insertions mapped to chromosome 2 at an average insertion gap of 14 bp. This high density of insertions across both chromosomes indicated that the level of saturation within the B. pseudomallei genome was sufficient to identify essential genes by a negative selection screen (Seyednasrollah et al., 2015).

We anticipated that a lower number or even a lack of transposon insertions (TIS) would be detected in regions that contain essential genes. This is because the disruption of an essential gene would render the insertion mutant bacteria non-viable and thus, most likely to be absent from the constructed mutant library. However, there are essential genes that can still tolerate disruptions at the extreme 3’ and 5’ ends of the open reading frames (Christen et al., 2011; Le Breton et al., 2015). To circumvent the likelihood of these non-disruptive insertions in the 5’ and 3’ ends, we limited our analysis to only TIS located within the sequence spanning the 5%–90% portion of each locus. Next, we normalized the number of TIS identified per gene based on the length of the respective gene to calculate the insertion index of each gene. The insertion indexes were fitted into a bimodal distribution and we obtained two peaks corresponding to essential and non-essential gene sets ( Supplementary Figure S3 ). To determine the essentiality cutoff point, we assumed a log2-likelihood ratio of less than -2. Thus, if the insertion index value of a gene is less than the cutoff value, a gene is at least four times more likely to be essential. We avoided any misinterpretation of a gene as “essential” by discarding sequence reads which mapped to more than one region. Using this premise, 79 genes (of which 40% were insertion sequence (IS) and putative transposases) were excluded from the essential list. Based on these strict selection criteria, a list of 492 genes was obtained and these genes are predicted to be essential for B. pseudomallei survival on LB agar supplemented with tetracycline ( Supplementary Data Table 1 ).

There was a distinct chromosome-level segregation whereby approximately 90% of the 492 genes were located on chromosome 1 ( Figure 1 ). This observation is in agreement with the B. cenocepacia TraDIS-derived essential data set (Wong et al., 2016) where the B. cenocepacia essential genes encoding for core cellular function are also enriched in Chromosome 1. Nonetheless, the lower number of essential genes identified in the B. pseudomallei chromosome 2 could be attributed to the distinct essentiality cutoffs applied to each chromosome to offset the imbalanced distribution of TIS. As the average gap between each TIS (14 bp) was larger in Chromosome 2, this meant a potentially higher rate of false discovery in chromosome 2 due to an increased possibility of genes being unrepresented in the transposon library by random chance. The number of genes that can be labelled as “essential” will be reduced with the application of lower essentiality cutoffs (DeJesus et al., 2017).

The identified B. pseudomallei R15 essential gene candidates were categorized into their functional classes based on COG (Clusters of Orthologous Groups) designations ( Figure 2 ). The distribution of essential genes within each functional category was similar to other different bacteria including B. cenocepacia (Luo et al., 2014; Wong et al., 2016) and also the synthetic Mycoplasma mycoides JCVI-syn3.0 (Hutchison et al., 2016). Up to 25% of the essential genes encode proteins required for fundamental biological processes necessary to sustain cellular life e.g., enzymes involved in DNA replication, transcription and protein translation. About 7% of the essential candidates fall in the COG functional category “cell wall/membrane/envelope biogenesis”, highlighting the importance of maintaining cell structure and integrity for bacterial viability. In addition, genes involved in metabolism are also important for bacterial survival, with about 10% and 8% of the essential genes in the categories of “coenzyme transport and metabolism” and “energy production and conversion”, respectively. As expected, about 20% of the genes were annotated as encoding for proteins of unknown function or hypothetical proteins.

Among the 492 genes predicted as essential in B. pseudomallei R15, 423 genes have homologs in the B. pseudomallei reference strain K96243. In a previous TraDIS study, a set of 505 genes were predicted to be essential in B. pseudomallei K96243 and these genes and their encoded proteins were proposed as excellent targets for new antimicrobials (Moule et al., 2014). Of these identified K96243 essential genes, 243 were also found to be essential in R15 and this set of genes most likely represents common essential genes shared by B. pseudomallei strains. These common core essential genes are enriched in COG categories such as “cell wall/membrane/envelope biogenesis”, “translation, ribosomal structure and biogenesis” and “energy production and conversion” ( Supplementary Table S1 ).

Many of the genes located in genomic island 8 (GI8; bpsl1637 – bpsl1709) have previously been predicted as essential in K96243 (Moule et al., 2014). GI8 contains transport proteins, haemolysin-related proteins, amine catabolism proteins, various regulators and YadA-like exported protein (Holden et al., 2004). Interestingly, in R15, homologs of these K96243 genes located within GI8 (r15_1836 − r15_1871) were found to be non-essential. In the R15 genome, these homologous genes are separately located in two genomic islands, GI05 and GI06. As shown in Figure 3A , a high density of transposon insertions can be observed in this genomic region (indicated by the grey arrow), implicating that disruption of these GI-associated genes did not impact R15 viability. There is an additional set of R15 essential genes (59) with no homolog in K96243 and hence, is unique to the R15 strain. However, as most of these R15-specific essential genes encode hypothetical proteins, their role(s) in maintaining R15 viability remains to be elucidated. Of particular interest, gene r15_2362, which encodes the virulence-associated E family protein VapE, was predicted to be essential, as well as its neighboring genes r15_2363 (hypothetical protein), r15_2364 (hypothetical protein), r15_2365 (hypothetical protein) and r15_2366 (repressor) ( Figure 3B ). Intriguingly, the VapE gene is also present in other B. pseudomallei strains including 1026b, a clinical strain isolated from a septicaemic melioidosis patient blood sample (Johnson et al., 2015). The presence of repressor and terminase encoding genes in this VapE-containing region suggests that this genomic region is likely to be prophage-related. This hypothesis is further supported by the absence of R15_2332 − R15_2376 homologs (~ 40 kb) in K96243, suggesting that this genomic region is not evolutionary conserved but most likely, horizontally-acquired.

B. thailandensis (Baugh et al., 2013), B. pseudomallei K96243 (Moule et al., 2014) and B. cenocepacia (Wong et al., 2016) essential genes were compared to determine the core essential gene set for the Burkholderia genus as identified using TraDIS and Tn-Seq. Our TraDIS-derived data set for B. pseudomallei R15 was combined with that of K96243 (Moule et al., 2014) into a final repertoire of 685 genes and is hence forth collectively referred to as B. pseudomallei essential genes set.

Of the 685 B. pseudomallei essential genes, 505 B. cenocepacia orthologs were noted but only 294/505 of these B. cenocepacia orthologs were previously determined to be essential in B. cenocepacia (Wong et al., 2016). When the set of genes was compared with the 406 B. thailandensis essential genes (Baugh et al., 2013), we identified 392 B. pseudomallei orthologs of which 340 (86%) were deemed essential in the B. pseudomallei data set. Taken together, a total of 258 common essential genes were present in B. pseudomallei, B. cenocepacia and B. thailandensis, and most likely represent the core essential genome shared by many species of the Burkholderia genus ( Figure 3C ; Supplementary Table 2 ). Not surprisingly, these genes are enriched in the functional categories of “transcription”, “translation, ribosomal structure and biogenesis”, “replication, recombination and repair”, “cell wall/membrane/envelope biogenesis” as well as “energy production and conversion”.

B. pseudomallei can thrive in various environmental conditions including nutrient-depleted water (Inglis and Sagripanti, 2006). Identifying genes that are necessary for bacterial growth under a defined condition can reveal the lifestyle adopted by the bacteria. To demarcate the subset of genes which are essential for B. pseudomallei growth in nutrient-rich (LB) and nutrient-depleted (M9) media, the input mutant pool was grown in LB and M9 minimal medium over three consecutive passages. At each passage, individual output pools were harvested and subjected to TraDIS analysis to detect any loss of mutants from the initial input pool. This approach allowed mutant selection over continued in vitro growth in either the nutrient-rich or nutrient-depleted media. We predicted that mutants of essential genes would be absent in the initial input pool while mutants with reduced fitness (conditionally essential) would be lost only during in vitro passages. A summary of TraDIS sequencing results for the in vitro grown output pools is shown in Table 1 .

In this study, we defined genes which are essential for in vitro growth as genes for which insertions resulted in fitness defects (i.e., reduced growth rates) and not complete growth arrest as defined for the input pool. Mutants carrying mutations that confer fitness disadvantages would most likely be outcompeted within the mixed population and under-represented after several consecutive passages.

The consensus sets of genes essential for in vitro growth was obtained based on genes that were identified as essential in at least two of the three pools of the same media. This approach generated two sets of data corresponding to LB (607 genes) and M9 (590 genes), respectively and this number of genes includes the essential genes predicted for the input pool. We then compared the three different data sets (initial input pool, LB and M9) ( Figure 4A ). There were 368 genes which overlapped and were deemed essential for all conditions and a total of 105 genes were categorized as “general essential genes” critical for bacterial growth in vitro as all were conditionally essential under both LB and M9 growth conditions. An additional set of 93 genes is required for B. pseudomallei growth in LB but not M9; conversely, 96 genes are conditionally essential when bacteria are grown in nutrient depleted M9 but not LB. These genes were hence classified as “condition-specific” ( Supplementary Data Table 3 ). The “general, LB-specific and M9-specific” essential genes locations within the B. pseudomallei genome are shown in Figure 1 .

B. pseudomallei wild type transposon mutant pools were used as the input agents to infect C. elegans. The input pools were initially cultured in LB medium and then spread on Nematode Growth agar (Tan et al., 1999). As mentioned earlier, the multi-step process to prepare the bacteria could have introduced selection pressure on the mutant population. Hence, there was a possibility that some mutants were already lost from the initial input pool before this pool was subjected to any in vivo selection. We assessed for this loss by analyzing mutants grown in LB and NGM which are referred to as output-LB and output-NGM (in vitro output pools).

A large-scale infection assay was performed using thousands of worms per experiment. Intestinal mutants were recovered from a population of worms at 6 hrs and 24 hrs post-infection (hpi). The infection experiments were conducted separately for a total of four times, and a mutant pool of approximately 1 x 10⁷ mutants were obtained for each in vivo output pool (6 hpi and 24 hpi). TraDIS libraries of in vitro and in vivo output pools were sequenced with an average of 25 million reads generated for each library of which, > 94% of the total reads contained the correct 10 bp transposon tag. Over 62% of the tagged reads were uniquely mapped to the B. pseudomallei R15 genome, resulting in 480,000 - 690,000 unique TIS identified across both chromosomes ( Table 2 ). For output-6 hpi and output-24 hpi, one TIS was detected every 14 bp and > 50 reads were available for each insertion. This high-density genome saturation is sufficient to estimate the relative abundance of mutants before (input pool) and after in vivo selection (output pool). A gene was important or costly for bacterial fitness in the infected host if the abundance of relative reads was at least four times lower in the in vivo output pool (log₂ fold change< -2; adjusted p-value<0.05) than that in the initial input pool.

One hundred and thirty genes from output-6 hpi and 180 genes from output-24 hpi met the criterion of log₂ fold< than -2. The M9-, LB- and NGM-specific conditionally essential genes required for growth in vitro were pooled and removed from the list of genes required for bacterial survival in vivo. Therefore, 98 genes were required for B. pseudomallei survival in the C. elegans intestinal lumen during early infection (6 hours post-infection) while 130 genes were deemed important for survival up to 24 hpi. Of the two sets of genes, 84 genes were important for B. pseudomallei in vivo fitness throughout the worm infection period (6 and 24 hours post-infection) ( Supplementary Data Table 4 ). Genes that negatively impacted B. pseudomallei survival in vivo were classified based on the COG annotation system. As depicted in Figure 5 , about 50% of the genes encode poorly characterized proteins with no assigned COG function, hence, how the encoded proteins contribute towards B. pseudomallei fitness in the infected host remains unclear. Approximately 30% of the genes were categorised as “cell wall/membrane/envelope biogenesis”, “transcription”, “carbohydrate transport and metabolism” as well as “defence mechanisms”.

To validate the TraDIS-derived in vivo data, mutants of two selected genes (bpsl2988 and bpsl3313) were obtained. These two genes are deemed essential for in vivo survival at both 6 hpi and 24 hpi, with the numbers of unique TIS significantly reduced by more than 60-fold when compared with the input pool ( Table 3 ). bpsl3313 encodes a H-NS-like transcriptional regulator while bpsl2988 encodes a putative sugar kinase. In a previous study, we screened a library of B. pseudomallei R15 Tn5 transposon mutants in a C. elegans infection model and identified bpsl2988 as one of the genes with significant attenuation in worm killing as compared to the wild type (data not published). The ability of the single-gene deletion mutants (Δbpsl3313) and Tn5 insertion mutant (bpsl2988::Tn5) to survive and proliferate in the C. elegans intestine was measured by enumerating the number of in vivo colonizing bacteria at 6 hours and 24 hours post-infection.

As shown in Figure 6 , during early infection (6 hpi), intestinal bacterial loads in worms infected with Δbpsl3313 and bpsl2988::Tn5 mutants were significantly lower (t-test, p<0.05) as compared to the wild type (WT) strain. After prolonged infection (24 hours), only bpsl2988::Tn5 showed limited intestinal colonization (t-test, p<0.01), whilst the ability of Δbpsl3313 to survive in the worm intestine was comparable to that of the wild type ( Figure 6 ). Insertions in these two genes negatively impacted mutant fitness in the worm intestine at both 6 hpi and 24 hpi, with relative mutant abundance being significantly reduced up to 65-fold. As TraDIS analyzed a pool of mixed mutants grown in competition, reduction in mutant survival by a similar degree may not be reproduced in an experiment that involved infection with only a single strain. A decrease in bacterial survival in vivo may be associated with a general growth defect. When grown in LB and M9 minimal media, there was no significant difference in growth rate observed for the two mutant strains as compared to the wild type (WT) ( Supplementary Figure S4 ), thus confirming that reduced intestinal colonization by Δbpsl3313 (at 6 hpi) and bpsl2988::Tn5 (at both 6 and 24 hpi) was not due to defects in bacterial growth or enhanced sensitivity to limited nutrient availability.

The ability of Δbpsl3313 and bpsl2988::Tn5 mutants to kill the nematode was evaluated in a standard slow-killing assay. The mean time to death (TD_mean; the time required to kill 50% of the worm population) of infected young worms was determined. bpsl2988::Tn5 was strongly attenuated ( Figure 7 ) with a TD_mean )of 41.07 ± 0.72 hours compared to the wild type (TD_mean 32.47 ± 0.62 hours) (Log-rank (Mantel-Cox) test, p<0.0001). Surprisingly, the Δbpsl3313 mutant was more virulent with a significantly shorter TD_mean of 28.26 ± 0.39 hours compared to the wild type (Log-rank (Mantel-Cox) test, p<0.0001).

As different model hosts respond to B. pseudomallei infection differently (Lee et al., 2011), the well-established BALB/c mouse model for melioidosis was challenged with wild type B. pseudomallei, Δbpsl3313 and bpsl2988::Tn5 mutants. Mice survival was monitored over time. Unlike the results obtained in a C. elegans model, where Δbpsl3313 and bpsl2988::Tn5 exhibited different levels of virulence, these two mutants did not show any altered phenotype in mice. More than 60% of the mice infected with either mutants or wild type B. pseudomallei succumbed to the disease on day 2 ( Figure 8 ). This conflicting response between mice and C. elegans may be attributed to distinct mechanisms of host-pathogen interaction between B. pseudomallei and both animal models and different hosts’ immune response. Nonetheless, while retaining full virulence, bpsl2988::Tn5 could not colonize the organs of infected mice to the same extent as wild type B. pseudomallei. Bacterial burden in the organs of bpsl2988::Tn5-infected mice were significantly reduced, with the most pronounced colonization defect observed in the lung (t-test, p<0.0001) ( Figure 9 ). Conversely, Δbpsl3313 did not show any reduced colonization in the infected mice organs as compared to the wild type. Again, this highlights the important role of BPSL2988 in B. pseudomallei survival in vivo.