P. aeruginosa PAO1 strain H103 (Zhang et al., 2000) was used as a WT control and for Tn-Seq studies. The Tn-insertion mutants used for susceptibility and growth defect validations were taken from an ordered Tn-insertion library in P. aeruginosa PAO1 (harboring Tn5 IS50L derivative Tn-insertions ISlacZ/hah and ISphoA/hah; Jacobs et al., 2003). E. coli strain SM10 λpir was used as a donor strain in Tn-Seq pool generation.

Media included lysogeny broth (LB), experimental control condition MHB, RPMI-1640 supplemented with 5% MHB (referred to as RPMI), RPMI-1640 supplemented with 5% MHB and 20% human serum pooled from anonymous donors under ethics approval [certificate number H04-70232]. RPMI and RPMI/serum were designed to be physiologically-relevant to wound exudate or blood and their preparation was described in detail previously (Belanger and Hancock, 2021).

The pBT20 plasmid was used for generation of the Tn-Seq pool and was obtained from the Filloux lab, Imperial College, London (Kulasekara et al., 2005). This is a mariner Tn vector with the Himar1 Tn that randomly inserts into Thymine-Adenine (TA) sites within the genome. The plasmid has the following characteristics: Δbla Tel r; The Tel r cassette, consisting of the kilA and telAB genes, is flanked by two identical FRT sequences. Primers used in this study are summarized in Supplementary Table S1.

A Tn-Seq pool was constructed in P. aeruginosa strain PAO1 using the vector pBT20. First, E. coli SM10λpir + pBT20 was grown overnight at 37°C on LB agar containing 100 μg/ml ampicillin (Sigma-Aldrich). P. aeruginosa PAO1 was grown at 42°C overnight on LB agar with no antibiotic. The bacteria were scraped off the plates and mixed in equal volumes of each at a ratio of OD₆₀₀ 40:20 donor (E. coli): recipient (P. aeruginosa). Conjugation mixtures were spotted on to LB agar plates and incubated for 2 h at 37°C. All conjugation spots were then scraped into LB and plated on LB agar with 25 μg/ml irgasan (Sigma-Aldrich) and 25 μg/ml gentamicin (Sigma-Aldrich). The cells were grown overnight then counted and collected in LB, and stored in 20% glycerol at −80°C. The procedure was repeated until at least 200,000 mutants were collected. The collected mutants were thawed at 37°C, normalized to equalize the number of colonies per mL per conjugation, and then pooled and recovered at 37°C. They were recovered by adding 0.5 ml of pooled cells into 50 ml LB with 15 μg/ml gentamicin. 20% glycerol was added to the final pool and it was aliquoted into 1 ml volumes and flash frozen with liquid nitrogen before storing at −80°C.

To determine which mutants existed in the Tn-Seq pool at baseline and after any selective pressure, total genomic DNA was extracted from the samples and the regions adjacent to the Tn-insertions were amplified and sequenced. Primers used for Tn-Seq are summarized in Supplementary Table S1. First, total DNA was extracted using the DNeasy Blood and Tissue Kit (QIAGEN). The concentration and purity of the DNA was measured using a NanoDrop 2000 (Thermo Fisher). Genomic DNA was used in four replicate PCR reactions to amplify the Tn-genome junctions via two nested PCR steps which are described in detail in Supplementary Materials. Products of replicate reactions of the second PCR step were pooled together and two-sided size selection was performed with Agencourt AMPure XP magnetic beads. DNA from each pool was eluted into 40 μl total in nuclease free Tris-EDTA buffer. Four aliquots of 5 μl were used for indexing and sequencing. The indexing PCR was done according to the Illumina protocol for amplicon sequencing using Nextera Index Kit v2 SetA (Illumina). Replicates were pooled and a final PCR cleanup was performed with Agencourt AMPure XP magnetic beads with elution into 40 μl nuclease free TE buffer. Libraries were quantified using Quant-iT dsDNA Assay Kit (Thermo Fisher) analyzed on the Bioanalyzer 2,100 (Agilent Genomics). Samples were diluted to 4 nM and pooled with other samples for an HiSeq sequencing lane, at 3–6% of Tn-Seq library total per lane. Samples were run on a HiSeq 2,500 single end 100 bp run (Illumina).

The quality of raw sequence reads was assessed using FastQC (v0.11.6; Andrews, 2010) and MultiQC (v1.6; Ewels et al., 2016). Conditionally essential genes were determined using two, complementary bioinformatics tools: TraDIS v1.4.5 (referred to as Tradis; Barquist et al., 2016) and TRANSIT v2.0.0 (Transit; DeJesus et al., 2015). The sequence reads were aligned against P. aeruginosa PAO1 reference genome from the Pseudomonas Genome Database v17.1 (accession GCF_000006765.1; PseudoCap version 138). Tradis was run using default parameters by aligning reads against the reference genome using SMALT and determining insertion counts per gene. Transit used TPP pre-processing tool to map raw reads against the reference with BWA v0.7.17 (Li and Durbin, 2009) with default parameters used to tabulate the counts of reads per TA sites. The counts were passed through the Gumbel method of Transit for calculating the probability of essentiality of each gene using a Bayesian model. Genes predicted to be essential by Gumbel or Tradis in at least 4 out of 5 replicates (in experimental replicates) or 2 out of 3 replicates (for the T0 pool) were compiled into a final list of essential genes.

Statistical significance of overlap between the Tn-Seq pool selected in different physiologically-relevant media was performed using hypergeometric distribution calculations and Fisher’s exact test with the GeneOverlap v1.32.0 package from BioConductor release 3.15 (Shen and Sinai, 2022) in R v4.1.2 (R Core Team, 2021). Overlaps in essential genes found between two physiologically-relevant media conditions was considered significant with an odds ratio greater than 1 if p < 0.05.

Functional class enrichment was performed using functional class annotations from PseudoCap (Winsor et al., 2016) and hypergeometric distribution calculations were done using phyper from the stats package v4.3.0 in R (Johnson et al., 1992; R Core Team, 2021) and correcting for multiple testing using false discovery rate (FDR) corrections. Functional classes considered significantly enriched within a comparison were those with adjusted p value <0.05.

Tn-insertion pools were inoculated at 1 × 10⁷ CFU/ml in 25 ml cultures in MHB, RPMI and RPMI/serum in five separate biological replicates performed over multiple days. The cultures were grown to 5 × 10⁸ CFU/ml at 37°C and bacteria were harvested from 1 ml each. Total genomic DNA was extracted, libraries were prepared and sequenced, and essentiality analysis was performed using Tradis and Transit. Conditionally essential genes were compared between media (physiologically-relevant conditions vs. MHB), and essential genes in MHB were removed from the physiologically-relevant media gene lists before comparing to other host mimicking conditions.

Validation experiments were performed with individual Tn-insertion mutants grown in physiologically-relevant in vitro media in 96-well plates. The wells containing RPMI/serum or MHB were inoculated with minimum 3 replicates of each mutant in each medium at 1 × 10⁷ CFU/ml and grown for 16 h in a plate reader (BioTek) with constant shaking and heating to 37°C. Select mutants with interesting growth defects were then grown in 25 ml flask cultures in RPMI/serum. Cultures were grown at 37C for 25 h and samples were taken and plated for CFU every 3 h for the first 9 h and then at the final time point. Growth cultures were performed with PAO1 WT grown in parallel to mutant strains.

Animal experiments were performed following Canadian Council on Animal Care (CCAC) guidelines and were approved by The University of British Columbia Animal Care Committee [certificate number A14-0363]. The mice used in this study were seven-week-old female outbred CD-1, purchased from Charles River Laboratories, Inc., (Wilmington, MA). The average weight was about 25 ± 3 g at the time of the experiments. One to 3% isoflurane was used to anesthetize the mice. Mice were euthanized with carbon dioxide.

For murine abscess Tn-Seq an aliquot of the PAO1 Tn-Seq pool was grown to mid log phase, washed, and resuspended in PBS to and OD₆₀₀ of 0.5. The pool was administered to mice subcutaneously with a total bacterial inoculum of 2 × 10⁷ CFU. Abscesses were allowed to form for 18 h, at which time the mice were sacrificed, and the abscesses removed. The experiment was repeated with five replicates.

Enzymatic digestion of the abscess tissue and separation of bacterial cells followed a previously published protocol with modifications (Garcia-Garcerà et al., 2013). The abscess tissue was homogenized in 1 mL sterile PBS with sterile glass beads and 10 μl were removed for CFU counting. The tissue was then enzymatically digested at 37°C for 45 min, inactivated and filtered. The final, filtered, washed cells were pelleted and used for DNA extraction using the same method as for Tn-Seq in host mimicking in vitro conditions. Library prep, sequencing and essentiality analysis was performed as described above.

Survival of ordered Tn-insertion mutants for genes identified through Tn-Seq in the murine abscess model was performed in a similar fashion to Tn-Seq experiments.

Tn-Seq was also performed in a human skin in vivo-like model (Wu et al., 2021). The human skin organoid (skin) model used N/TERT keratinocyte cells cultured on filter inserts in 12-well plates containing growth factors and tissue culture medium. Over approximately 3 weeks, the cells grew, differentiated, and stratified into a dermal-like tissue on which bacteria could be inoculated.

The Tn pool was grown in LB to mid log phase, washed with PBS, and inoculated onto the skin at 1 × 10⁶ CFU in five μL of PBS and allowed to develop for 48 h. The skin was then washed with PBS, and the entire filter with skin and bacteria attached was removed from the tissue culture wells, homogenized, and plated to measure survival. The skin was then digested following the same protocol as for abscess tissue. Whole genomic DNA extraction was performed and the Tn-genome junctions were amplified and sequenced as above. The experiment was repeated with four replicates from different skin batches.

Competition between PAO1 ordered Tn-insertion mutants for genes identified through Tn-Seq and wild type (WT) PAO1 was measured after 48 h of bacterial growth on the skin. A control Tn-insertion mutant was included where the insertion is at the end of the general essential gene dnaG and thus does not demonstrate an insertional effect and would not be expected to have a growth advantage or disadvantage when compared to WT. The skin was inoculated with 1 × 10⁶ CFU in 5 μl of PBS containing 50% WT and 50% mutant population. The infection was left for 48 h, then the skins were washed with sterile PBS, homogenized using glass beads, and plated on LB agar (for growth of all bacteria) and tetracycline (50 μg/ml: for growth of Tn-insertion mutants). The ratios of mutant to WT were measured for all mutants tested (Macho et al., 2007). Competitive fitness indices (CI) were assessed using the following formula CI = (M_O/WT_O)/(M_I/WT_I), where, M was the CFU/mL of the mutant, WT was the CFU/mL of the wild type, subscript O is the output counts after a biofilm was established on the skin, and subscript I was the input counts in the inoculum.

The web-based application PaintDB (Castillo-Arnemann et al., 2021) was utilized to create networks in order to visualize RNA-Seq and Tn-Seq data together. To explore genes essential under physiologically-relevant conditions and the transcriptional patterns of genes in the same pathways, RNA-Seq data for P. aeruginosa grown in RPMI/serum (Belanger et al., 2020) was combined with the gene list of genes conditionally essential in RPMI and/or RPMI/serum. Subnetworks of significantly enriched pathways were exported and edited for visualization using Cytoscape (v 3.8.2; Shannon, 2003). Using the ontologies tool in PaIntDB, subnetworks were constructed for pathways that were significantly enriched in these datasets using Fischer’s exact test with Bonjamini/Hochberg correction and a 0.05 significance threshold.

A Tn-Seq pool was generated in P. aeruginosa strain PAO1 grown in LB, using the mariner Tn Himar1 that randomly inserts into TA sites within the genome (Lampe et al., 1998). Over 200,000 mutants were collected, to generate a Tn pool that was more than two-fold larger than the possible insertion sites in the genome. To identify essential genes in LB, before expansion experiments were performed, three individual aliquots (biological replicates) of the Tn-insertion pool (denoted as T0) were sequenced. On average 4.4 million reads mapped to the genome, with between 90,000 and 100,000 unique insertions sites (Supplementary Dataset 1). Nineteen genes were excluded from the analysis as they contained no TA sites (Supplementary Dataset 1). To determine essential genes, a pipeline was constructed to utilize the analysis applications Tradis (Barquist et al., 2016), which determined essentiality by assessing insertions per gene, and Transit (DeJesus et al., 2015), which calculated the probability of essentiality of each gene using reads per TA site and a Bayesian model. By integrating two distinct analysis tools, the pipeline took into account variation in results that might occur due to differing algorithms in an attempt to reduce the risk of reproducibility issues. Genes were termed essential if they were identified as essential by either tool in at least 2 out of 3 replicate aliquots. A total of 654 genes were predicted as essential in the T0 pool in this study.

To analyze the robustness of the Tn-Seq pool, the essential genes identified in our experiment were compared to essential genes identified in previous Tn-Seq and ordered Tn mutant studies in P. aeruginosa, to assess similarities between the libraries (Table 1; Supplementary Dataset 1). Previously, investigations of P. aeruginosa essentiality have predicted between 336 and 1,394 generally essential genes when this organism is grown under rich growth media conditions (Table 1; Liberati et al., 2006; Skurnik et al., 2013; Lee et al., 2015; Turner et al., 2015; Poulsen et al., 2019). Overall, there were 147 genes predicted as essential in rich medium in our T0 pool, as well as in all six published essentiality studies (Supplementary Dataset 1). In addition to those predicted in all studies, another 135 were essential in the T0 pool and in five other essentiality studies, for a total of 282 common, frequently identified essential genes. Using hypergeometric distribution statistics, we determined that our T0 essential genes had significant overlap with all previous studies examined, with the highest correlation with Lee et al. (2015), Poulsen et al. (2019), and Turner et al. (2015) (Supplementary Dataset 1). Additionally, we found 10/26 Pseudocap functional classes (Winsor et al., 2016) showed common significant enrichment of essential genes in five, six, or seven of the studies examined here (Supplementary Table S2). These results indicate that the bacterial growth condition or method used to determine essentiality might strongly impact on the number and categories of genes identified as essential. Despite this, we determined significant overlapping functional classes to previous studies, and found our T0 pool to be a good representative of generally essential genes of P. aeruginosa in nutrient rich LB.

Following the characterization of the PAO1 Tn-insertion pool, this library was used to determine which genes were conditionally important for survival in physiologically-relevant in vitro media (Belanger et al., 2020), in an in vivo murine abscess model (Pletzer et al., 2017b), and in an in vivo-like human skin organoid model (Wu et al., 2021). From this, we aimed to understand whether genes required for survival of P. aeruginosa in host mimicking media were also required in established abscess and skin models, and to discover unexplored pathways that might be important for survival in wound models. By focusing on genes determined to be conditionally required in each medium (mutants completely absent after expansion) and excluding genes contributing to fitness but not absolutely required (mutants with decreasing numbers), we identified genes that were likely impacting to the greatest extent on survival and fitness in the host environment.

Comparing predicted essential genes lists between conditions, we found there were 67 genes identified as conditionally important for survival in both abscess and the skin model that were not identified in RPMI or RPMI/serum. Interestingly, this included a larger number of genes involved in adaptation and protection, chaperones and heat-shock proteins, transcription, translation and DNA repair (Figure 1B; Supplementary Table S4). Among the genes involved in adaptation and protection that were found only in the murine abscess and skin models were the type VI secretion system gene tsi2, genes encoding the soluble bacteriocin pyocin S4, chemotaxis genes, and Psl polysaccharide synthesis gene pslI.

Alternatively, there were 80 genes required for survival of P. aeruginosa in physiologically-relevant in vitro media only, many of which belonged to functional classes that were also predicted as essential in vivo, e.g., those influencing energy metabolism, amino acid and nucleotide metabolism, membrane integrity, and iron acquisition (Supplementary Table S5; Figure 1B). It is likely these functional classes are not discriminatory and are required for both in vitro and in vivo/in vivo-like conditions, but any differences might reflect differences in the wiring of metabolic pathways.

These data supported that genes identified in physiologically-relevant in vitro media provide a good representation of conditionally important genes in the host but likely did not provide an exhaustive list of all essential functions, and therefore has limitations.

To determine trends in P. aeruginosa gene requirements in the utilized infection models, we compared conditionally required genes identified in vitro, in murine abscess, and in human skin models and found 110 shared genes between the three conditions when excluding genes essential in MHB (Figure 1; Supplementary Dataset 2). Classification of these 110 genes into Pseudocap functional classes identified membrane-related functions (14 genes), amino acid metabolism (8), biosynthesis of cofactors and prosthetic groups (4), nucleotide biosynthesis and metabolism (6), transcriptional regulation (6), and transport of small molecules (11) as the most represented (Supplementary Dataset 2). Genes in these functional classes were explored in more detail below.

Previous studies examining genes important for survival of P. aeruginosa grown under physiologically relevant growth conditions have indicated that virulence, secretion and iron acquisition are important factors involved in survival and infection (Reimmann et al., 2001; Galle et al., 2012; Turner et al., 2014; Lee et al., 2015; Turner et al., 2015; Poulsen et al., 2019). We found that several genes encoding proteins involved in transport, secretion and membrane integrity were predicted as essential in all physiologically-relevant conditions in this study (Table 2). This included genes important for transport of divalent cations and metallic cations such as Fe³⁺, e.g., adjacent genes encoding the pyochelin synthesis proteins, PchG and PchF (Supplementary Dataset 2). Identifying genes involved in iron uptake is noteworthy, since P. aeruginosa has multiple systems for synthesizing siderophores, which are be diffusible molecules that bind and uptake iron. For this reason, we would expect the mutants to be complemented in trans. However, in our models, we found pyochelin synthesis mutants in the Tn-Seq pools did not survive. This being said, iron acquisition genes have also been identified in previous Tn-Seq studies in murine chronic wound models (Supplementary Table S6; Turner et al., 2014) and it is well known that they play an important role in infection and virulence.

Other genes involved in membrane functions and transport of small molecules, that were predicted as essential in host mimicking conditions in this study (Table 2; Supplementary Table S6) as well as previous studies included (a) genes encoding ferredoxin protein PA2297 and nitrate reductase NapF/NapE, which were also found to be essential in murine wound models (Turner et al., 2014), (b) sodium/hydrogen antiporter proteins (ShaDF) implicated as important in sodium homeostasis and virulence of P. aeruginosa in vivo (Kosono et al., 2005) and previously found to be essential in murine wounds, SCFM and lung sputum (Turner et al., 2014, 2015; Lee et al., 2015), (c) copper transport gene copA2 which was previously found to be essential in wound, bovine serum, and SCFM (Turner et al., 2014, 2015; Poulsen et al., 2019), as well as (d) translocase genes coding for the TatABC and Psc operons and HxcUW pseudopilins important for type 2 (T2SS) and type 3 (T3SS) secretion systems and previously implicated as essential in wound, bovine serum, and SCFM models (Turner et al., 2014, 2015; Morgan et al., 2019; Poulsen et al., 2019).

Since many virulence and transport genes had overlapping essentiality between this study and previous studies, we examined whether these genes also had similar transcriptional patterns in RPMI/serum to those observed in studies exploring physiologically-relevant conditions. To do this, previously published RNA-Seq data collected from P. aeruginosa grown under RPMI/serum cf. MHB (Belanger et al., 2020) were compared to qRT-PCR analysis in the murine abscess model, (Pletzer et al., 2017b), and transcriptomic data in chronic and acute murine wound infections (Supplementary Table S7; Turner et al., 2014). Both previous abscess/wound studies focused on pyoverdine and phenazine iron uptake pathways, regulation of T3SS and T2SS, alginate synthesis, Psl polysaccharide production, motility, and rhamnolipid production as markers for virulence in vivo. Transcriptomics of P. aeruginosa grown in vitro in RPMI/serum cf. MHB (Belanger et al., 2020) revealed increased expression of selected pyoverdine genes, and T2SS and T3SS genes, as was observed in vivo, but no alteration of expression of other genes involved in motility and rhamnolipid synthesis (Supplementary Table S7). This comparison implies similar regulatory patterns in secretion systems and virulence factors between host like conditions, but differences in the behavior of specific genes involved in motility and attachment. By observing essentiality and expression patterns of genes involved in virulence in both in vitro and in vivo host mimicking conditions compared to MHB, we found that some of the genes identified in this study had also been identified by previous studies (Lee et al., 2014; Turner et al., 2014, 2015; Pletzer et al., 2017b; Poulsen et al., 2019) in different host environments. These observations support that the host mimicking conditions used here, reflect a sufficient representation of genes considered essential for infection of P. aeruginosa in the host environment.

In addition to virulence factors and membrane transport genes, we found that a large proportion of classified essential genes in host mimicking conditions belonged to pathways involved in microbial metabolic functions. These genes were of interest since they might indicate altered metabolism and potential metabolic targets of P. aeruginosa in infection environments.

Amino acid metabolism genes that were required for survival in both abscess and skin models as well as physiologically-relevant in vitro media included genes encoding proteins for phenylalanine, tyrosine, lysine, histidine and methionine biosynthesis and utilization (Table 2; Klem and Davisson, 1993; Myers et al., 1993; Xie et al., 1999). The genes carA that encodes a subunit of carbamoyl-phosphate, involved in catalyzing the biosynthesis of a precursor for arginine and pyrimidines (Cunin et al., 1986), and nagZ that codes for a beta-N-acetyl-D-glucosaminidase involved in β-lactam resistance (Zamorano et al., 2010) were also required.

In this study, we also found that genes pyrF, purE, and carA belonging to pyrimidine and purine metabolism operons (Supplementary Figure S2), were required for survival in all media conditions (Table 2). Additionally, pyrD, purF, purH, and hypothetical pyrimidine biosynthesis gene PA3505 were predicted essential in 3 out of 4 host-like conditions. Genes belonging to purine and pyrimidine metabolism have been previously suggested to play an essential role in P. aeruginosa survival in host environments (Supplementary Table S6) such as blood (Samant et al., 2008), human serum (Weber et al., 2020), in vivo infection models (Turner et al., 2014), and in bovine serum (Poulsen et al., 2019) and SCFM (Turner et al., 2015).

Genes belonging to operons involved in carbon catabolism and biosynthesis of cofactors had interesting overlaps between in vivo and in vitro host mimicking conditions. Malonate decarboxylase genes mdcC and mdcE were required for survival in all physiologically-relevant conditions (Table 2) and were also previously identified as important in murine wound models (Turner et al., 2014). Genes involved in the biosynthesis of pyrroloquinoline quinone (PQQ) and cobalamin (Cob; Supplementary Figure S3) were required in some or all physiologically-relevant conditions. PQQ is a co-factor involved in energy metabolism and previously demonstrated to be induced during growth on ethanol, 1-propanol, 1,2-propanediol and 1-butanol (Gliese et al., 2010). Cobalamin, or vitamin B₁₂, is a complex cofactor containing a central chelated cobalt ion. CobC, cobD, cobO and cobQ were required in all physiologically-relevant conditions, and others such as cobE and cobH were essential in only certain physiologically-relevant media. Similar to iron acquisition, it was surprising to identify cobalamin biosynthesis as essential, since one would expect that loss of cobalamin biosynthesis would be complemented by other mutants in the pool. Nevertheless, the levels of cobalamin in human serum and RPMI are very low when compared to other vitamins (Supplementary Table S8), and this pathway has also been demonstrated as essential in other in vivo models (Turner et al., 2014), which could make it a limiting nutrient for survival in these conditions.

The availability of RNA-Seq data for P. aeruginosa grown under in vitro physiologically relevant conditions of RPMI and RPMI/serum (Belanger et al., 2020) gave us the ability to compare patterns of significantly enriched gene expression in these conditions, to the Tn-Seq data collected in this study. It was hypothesized that there would be overlapping enrichment in gene expression of pathways that were essential for survival of P. aeruginosa in the host environment. Conditionally essential genes in RPMI and/or RPMI/serum were integrated with RNA-Seq from P. aeruginosa grown in RPMI (Belanger et al., 2020), using a new web tool, PaIntDB, (Castillo-Arnemann et al., 2021) that maps such data to protein–protein interaction (PPI) networks in P. aeruginosa. PPI networks comprised nodes (circles) representing gene-encoded proteins connected by lines/edges that represent known or extrapolated PPIs representing physical, metabolic or regulatory interactions. Using the RNA-Seq and essential Tn-Seq genes, a zero-order PPI network of 1,921 genes/proteins was constructed, involving only direct interactions between these nodes., including 1,856 nodes from RNA-Seq data and 175 from Tn-Seq data. Using the ontologies tool in PaIntDB, subnetworks were constructed for pathways that were significantly enriched in these datasets. There were a total of 150 significantly enriched gene ontology (GO) terms found (p < 0.05).

A subnetwork was identified containing purine and pyrimidine byosinthesis and metabolism pathways with 76 enriched genes, 12 of which were conditionally essential in RPMI or RPMI/serum cf. MHB (Figure 2). The majority (10/12) of these essential genes were also observed to be essential in the murine abscess models, while five were essential in the skin model. Five genes were essential in both the abscess and skin models. Interestingly, most of the genes in these pathways were downregulated by up to 4-fold when compared to MHB, which was consistent with the findings of Turner et al. in murine chronic and acute wounds compared to MOPS-succinate (Turner et al., 2014). Although a smaller number of genes involved in biosynthesis of cofactors, prosthetic groups and carriers were indicated to be essential for survival in vivo, cobalamin biosynthesis was conditionally important and was also enriched in RNA-Seq data (Belanger et al., 2020) in host mimicking in vitro media (Figure 3). A subnetwork of 15 genes/proteins was obtained; of these, six were conditionally required in RPMI and/or RPMI/Serum, two in both abscess and skin, and one in only abscess or skin, respectively. Notably, cobD, cobQ, cobO, and cobC genes were essential for growth in multiple physiologically-relevant conditions tested, with cobD being found as essential in all conditions cf. MHB (Table 2). These genes were also downregulated under physiologically-relevant conditions when compared to MHB.

Therefore, integration of RNA-Seq and Tn-Seq data indicated the global importance of nucleotide and cobalamin synthesis pathways under physiological conditions, likely extenuated by the general downregulation of these pathways.

Since nucleotide metabolism was both essential and downregulated in physiologically relevant media, we hypothesized that a lack of precursor metabolites required for biosynthesis of nucleotides could explain the requirement and downregulated of genes from the nucleotide biosynthetic pathways under physiologically-relevant conditions when compared to MHB. We explored the available RNA-Seq data (Belanger et al., 2020) and found that genes required for the utilization of precursor metabolites for pyrimidines and purines were indeed upregulated (Supplementary Table S9). These included, the arcC (Baur et al., 1989) and ansB (Holcenberg et al., 1978) genes encoding elements of the arginine metabolic pathway catalyzing the production of carbamoyl phosphate from glutamine, which can then be used in pyrimidine metabolism, as well as glutamate catabolic genes pauABCD (Yao et al., 2011), and histidine catabolic genes hutGHIU which are involved in the production, from proteins, of amino acids previously proposed to be available in the host environment (Turner et al., 2014). Interestingly, genes involved in histidine metabolism were also essential under physiologically-relevant conditions, and it is possible that increased catabolism of arginine and histidine could be shuttled into increasing glutamine production, which could then be used in purine and pyrimidine metabolism. Similarly, an alternative gene proposed to be involved in pyrimidine synthesis, panE, was also upregulated under these conditions, providing a potential alternative path to pyrimidine synthesis in the host environment. These data are thus consistent with the importance of nucleotide biosynthesis in vivo and support that alternative pathways for nucleotide metabolism in the host environment might be utilized due to nutrient limitations in this environment.

To validate growth defects of selected essential genes belonging to the same pathways as genes identified as conditionally essential, we utilized ordered Tn-insertion mutants from the PAO1 library (Liberati et al., 2006). We selected all mutants for genes in nucleotide metabolism, and cobalamin biosynthesis that were found in the ordered Tn library. These included pyrC, pyrD, pyrE, pyrQ, pyrR, purK, and carA for nucleotide metabolism, and cobD, cobL, and cobP for cobalamin metabolism. In addition, iron acquisition gene pchF and secretion/toxin system genes exsE, toxA, and lipA were included to validate the importance of iron acquisition and virulence. Mutants were first tested for growth in physiologically-relevant in vitro media, cf. MHB in a 96 well microtitre growth set up (Figure 4). Growth defects in RPMI/serum were observed in 12 out of 16 mutants for genes involved in toxin secretion, and metabolism of nucleobases, vitamin B₁₂, and iron, with the largest defects observed for pyrD, pyrE, tadD, cobD, cobL and pchF. The remaining nine mutants had no growth defects in either medium. The lack of a growth defect despite conditional essentiality might be due to a competitive disadvantage in the Tn-Seq studies, cf. other mutants in the pool, that was not present when grown as isolated mutants.

Four mutants of interest were selected for more detailed assessment, including mutants for nucleobase metabolism genes pyrD::lacZ and pyrE::phoA, cobalamin metabolism gene cobD::phoA, and the negative regulator of T3SS exsE::lacZ as a virulence control. Growth kinetics were first investigated by inoculation of each of these mutants and WT PAO1 into media flasks at a starting concentration of 1 × 10⁶ CFU/ml. Cultures grown for 20 h at 37°C, and plated after 3, 6, 9 and 20 h to assess CFU/mL (Figure 5A). The exsE mutant showed no significant change in growth, while the pyrD, pyrE and cobD mutants had significant defects in growth in RPMI/serum, with 5.0 (p = 0.006), 3.9 (p = 0.005), and 1.7 (p = 0.02) fold less bacteria cf. WT, respectively.

Additionally, 2.5 × 10⁷ CFU of each mutant or WT were inoculated subdermally in mice to determine their ability to survive and form an abscess after 18 h (Figures 5B,C). The exsE mutant showed no growth defect and in fact demonstrated a slight but significant advantage in both CFU counts (2-fold more than WT, p = 0.03) and pathology (1.4-fold larger abscess than WT, p = 0.02). The pyrD and pyrE mutants demonstrated significantly decreased survival in vivo (2.8-fold, p = 0.02, and 29.4-fold, p = 0.005, less bacteria in abscesses cf. WT, respectively) as well as pathology, based on their ability to form abscesses (1.7 and 2.3-fold smaller abscess than WT for the pyrD (p = 0.0001) and pyrE (p = 7 × 10⁻⁵), mutants, respectively). Despite the fact that the cobD insertional mutant had only slightly decreased growth in vitro when compared to WT, in the murine model it exhibited 2.1-fold (p = 0.0008) and 7.9-fold (p = 0.006) decreases in abscess size and bacterial counts, respectively. To determine the effects that these mutations had on survival in the human skin organoid model, Tn mutants were mixed with WT PAO1 at a starting ratio of 1:1 and inoculated onto developed skin. The ratio of mutant to WT bacteria was used to assess the competitive index (CI) of the mutants after 48 h incubation on the skin (Figure 5D). A CI lower than one indicated a relative competitive fitness defect for that organism. As a negative control, we utilized a dnaG::ISlacZ/hah Tn mutant with an insertion at the end of this gene that did not disrupt the activity of this generally essential gene; it demonstrated no effect on growth rate in the skin organoid model with a CI of 0.82, that was not significantly different from WT (using a 1-sided t-test). The two mutants for nucleobase metabolism genes had the greatest growth defect in the human skin organoid model with CIs of 0.02 for pyrD::lacZ (p = 0.0002) and 0.01 for pyrE::phoA (p = 0.00005). The cobD mutant was also significantly reduced in growth in the skin model compared to WT with a CI of 0.21 (p = 0.004).