The F. tularensis strains used in this study included hypervirulent A.I strains SCHU S4 and MA00-2987, virulent A.II WY96-3418, and attenuated B strain LVS, and were all obtained from the Biodefense and Emerging Infections (BEI) Resources, which was established by the National Institute of Allergy and Infectious Diseases (NIAID). Select agent F. tularensis strains were transferred to the University of Nebraska Medical Center in Omaha following the requirements of the Federal Select Agent Program as outlined in the Animal and Plant Health Inspection Service/Centers for Disease Control and Prevention (CDC) Form 2, Guidance Document for Request to Transfer Select Agents and Toxin. Manipulation of viable culture material was performed by authorized individuals within a biosafety level 3 (BSL-3) laboratory certified for select agent work by the United States Department of Health and Human Services using laboratory biosafety criteria, according to requirements of the Federal Select Agent Program. For each experiment, F. tularensis strains were cultured from a master stock onto Remel Chocolate agar plates (Lenexa, KS) and incubated at 37°C with 5% CO₂ for 2 days before further subculturing or processing. Growth curves of F. tularensis were obtained by culturing in brain heart infusion broth (BHI) (Becton, Dickinson and Company, Sparks, MD) or Chamberlain’s CDM (Chamberlain, 1965) at 37°C with shaking in unbaffled flasks or in a Tecan Spark system in which OD₆₀₀ readings were recorded every hour.

For the RNA-Seq libraries, F. tularensis SCHU S4, MA00-2987, WY96-3418, and LVS were grown in BHI to mid-exponential growth phase in triplicate and in four independent experiments. Francisella tularensis cells were then treated with RNAprotect Bacteria Reagent (Invitrogen) to retain high RNA integrity. Next, FastRNA Pro Blue Kit (MP Biomedicals) or TRIzol (Invitrogen) was used to isolate total RNA, as recommended by the manufacturer. RNA was treated with Baseline-ZERO™ DNase (Epicentre Biotechnologies, San Diego, CA, United States) and the absence of residual genomic DNA was confirmed using control reactions without the addition of reverse transcriptase and subsequent PCR amplification with Francisella-specific primer, as previously described (Larson et al., 2011). To determine the RNA concentration, a Qubit fluorometer with the Qubit RNA HS Assay Kit (Invitrogen) were used, as described by the manufacturer. The integrity of the isolated RNA was checked by fractionation in an agarose gel containing ethidium bromide and visualization using a UV transilluminator.

The ScriptSeq Complete Kit for bacteria (Epicentre) was used to generate a ribodepleted, stranded transcriptome of coding and noncoding RNA for each of the four F. tularensis strains, specifically SCHU S4, MA00-2987, WY96-3418, and LVS. To monitor the ribosomal RNA removal process, RNA samples were assessed on a Fragment Analyzer (Agilent Technologies, Santa Clara, CA, United States). Twenty-four RNA-Seq libraries were prepared for each of the F. tularensis strains, totaling 96 RNA-Seq libraries for statistical power. Deep RNA sequencing of the F. tularensis libraries was performed by the University of Nebraska Sequencing Core facility using the Illumina NextSeq platform. Multiplexing of 48 samples per group and two groups in total were each ran on a single Mid-output Illumina flow cell, generating approximately 2.5 million 150 bp paired end reads per sample in 300 cycles.

In silico analysis of the RNA-Seq data was performed using genomic sequences for the aforementioned F. tularensis strains deposited in the National Institute of Health (NIH) GenBank database. The F. tularensis genome sequences were obtained from NCBI microbial genome database and had the following accession numbers: SCHU S4 (NC_006570.2), MA00-2987 (NZ_CP012372.1), WY96-3418 (NC_009257.1), WY-00W4114 (NZ_CP009753.1), FSC200 (NC_019551.1), and LVS (NC_07880.1).

Both FastQC and Trimmomatic were used for testing and adjusting the quality of the samples reads. FastQC was initially used for each read to provide quality data, similar to the average Phred score. The Phred score or quality (Q) score of a base is an integer value representing the estimated probability of an error (P) due to an incorrect base. Q and P were defined as follows:

Trimmomatic was utilized to remove the low-quality reads obtained by FastQC. Overrepresented sequences and an average Phred score of less than 20 per base (or 1% probability of an incorrect base) indicates low quality sequences. Therefore, the sequences with the Phred score of less than 20 or overrepresented sequences were eliminated from the reads. After the Trimmomatic modifications, reads with lengths of less than 35 bases were removed from the analysis, and only sequences 35 bases in length or longer were retained. To further ensure quality sequences, FastQC was again applied on the trimmed output results. To align the quality reads for each sample, the splice junction mapper TopHat2 was used (Kim et al., 2013). Bowtie 2 was utilized for indexing the genome assembly (Langmead and Salzberg, 2012). Cufflink identified expressed genes and calculated transcript levels in the aligned reads within the TopHat2 generated Sequence Alignment/Map (SAM) formatted files. The measurement of gene expression was determined using fragments per kilobase of exon per million (FPKM) reads, which quantifies the amount of data that matches a given coding sequence per kilobase length for a given gene or coding sequence per million reads analyzed (Croucher and Thomson, 2010; Giannoukos et al., 2012). The expression level of the same gene from the different F. tularensis strains were then compared to the normalized transcriptome data.

F. tularensis nucleotide and protein sequences of interest were obtained from the GenBank database. The Basic Local Alignment Search Tool (BLAST), which is provided by the NCBI within the National Library of Medicine (NLM), was used to identify regions of similarity between F. tularensis nucleotide or protein sequences. Nucleotide and protein alignments were performed using Clustal Omega (Sievers and Higgins, 2014). Metabolic pathways for gene products of interest were assessed using the Kyoto Encyclopedia of Genes and Genomes (KEGG) analysis tool. Note that as the NCBI annotations continue to evolve with intermittent iterations, speH may also be referred to as speD.

For reverse transcription quantitative real-time PCR (RT-qPCR), F. tularensis strains were grown in BHI or CDM as specified in triplicate and in three independent experiments. RNA was isolated, DNase treated, and checked for integrity and concentrations, as described above. To confirm the transcriptional expression of the genes of interest, first-strand cDNA was prepared using the SuperScript IV First-Strand Synthesis System and gene-specific primers, as recommended by the manufacturer (Invitrogen). Next, cDNA was amplified with appropriate gene-specific primer pairs. If conventional RT-qPCR was performed, products were evaluated by agarose gel fractionation, ethidium bromide staining, and densitometry, using ImageJ software (NIH). PowerUp SYBR Green Master Mix (Applied Biosystems) was also used with a QuantStudio 3 Real-time PCR System (Applied Biosystem), according to the manufacturer’s recommendations. Relative gene expression was determined by normalization to the internal control lpnA transcript and triplicate samples were evaluated in three independent experiments. Primers utilized for PCR and RT-qPCR are shown in Supplementary Table S1.

All oligonucleotide primers used in this study were synthesized by Thermo Fisher Scientific and are listed in Supplementary Table S1. Genomic DNA was isolated using the Gentra Puregene Cell Kit (Qiagen) or cetyl trimethylammonium bromide (CTAB), according to recommended procedures. Plasmid DNA was prepared with Zymoclean Gel DNA Recovery Kit (Zymo Research Corporation, Irvine, CA, United States), as recommended by the manufacturer. DNA restriction enzyme digests, cloning, and electrophoresis were performed according to standard protocols. PCR was performed using Platinum Taq DNA Polymerase High Fidelity enzyme and associated components (Invitrogen). Ligations were performed using a T4 DNA ligase (Roche) or NEBuilder HiFi DNA Assembly Master Mix (New England Biolabs, Ipswich, MA, United States). DNA fragments were purified using either a QIAquick PCR Purification or QIAquick Gel Extraction Kit (Qiagen). DNA concentrations were obtained using a Nanodrop spectrophotometer or Qubit fluorometer (Invitrogen). DNA sequencing was performed by the University of Nebraska Medical Center Genomics Core Facility using Sanger sequencing. Electroporation of plasmid DNA into F. tularensis or Escherichia coli was conducted as previously described (Horzempa et al., 2010), and an ECM 630 BTX Electroporation System (BTX Harvard Apparatus) was utilized for these procedures.

In-frame and markerless speE deletion mutants were generated in F. tularensis LVS and SCHU S4 using previously described procedures (Horzempa et al., 2010). To preserve operon integrity and any potential regulatory sites within the target gene sequence, nine consecutive codons at the beginning and end of speE were retained in the construct, along with three consecutive stop codons that were inserted after the last of the nine codons at the 5′ region of this gene. Fragments that include these regions and ~500 bp upstream and downstream of the speE coding sequence were amplified by PCR, digested, and ligated into PstI/SmaI digested pJH1. The resulting pJH1/speH_speA construct was DNA sequenced to confirm content and then transferred into F. tularensis LVS and SCHU S4 by tri-parental mating, as described previously (Horzempa et al., 2010). Merodiploid strains were recovered and transformed with pGUTS by electroporation for the expression of I-SceI (Horzempa et al., 2010). Colonies resistant to kanamycin were screened by PCR for deletion of the speE gene. To cure the F. tularensis ΔspeE mutant of pGUTS, the strains were passaged several times in tryptic soy broth containing cysteine, diluted, and plated on chocolate agar II plates, to obtain isolated CFUs. Plates were incubated for at least 3 days at 37°C with 5% CO₂ and colonies that formed were replica plated onto chocolate II agar plates with and without kanamycin. Colonies sensitive to kanamycin were isolated and retested for sensitivity to this antibiotic. Deletion of speE in the resulting F. tularensis LVS ΔspeE and SCHU S4 ΔspeE mutants were confirmed by PCR amplification of the relevant chromosomal locus and subsequent fractionation in an agarose gel and staining for comparison to the associated wild-type strain amplicon derived with the same primer pair. Content of this chromosomal locus in the F. tularensis LVS ΔspeE and SCHU S4 ΔspeE mutants was verified by DNA sequencing. RT-qPCR was used to confirm that speE was deleted in the ΔspeE mutant and that the adjacent genes were transcribed at levels similar to wild-type transcripts.

To complement the ΔspeE mutant, PCR amplification of the full-length speE gene was performed and the resulting amplicon was digested with NheI and BamHI. This PCR product was then ligated into pFNLTP, which was digested with the same restriction endonucleases. The groESL or native spe promoter were cloned upstream of speE into the KpnI and XhoI site of the pFNLTP plasmid containing speE. Both the promoter region and the speE gene in the pFNLTP expression plasmid were sequenced to verify content. The resulting pFN/Pr_groESL/speE and pFN/Pr_spe/speE expression constructs were then electroporated into the F. tularensis ΔspeE mutant for trans-complementation. Expression of speE in the complemented ΔspeE mutant was confirmed by RT-qPCR for triplicate samples in three independent experiments.

Bicinchoninic acid (BCA) assays were performed to determine the concentration of protein in the lysates obtained from F. tularensis SCHU S4, MA00-2987, WY96-3418, and LVS. Next, equivalent amounts of protein from the four different F. tularensis strains were reduced with DTT, alkylated with iodoacetamide, and digested overnight with sequencing-grade trypsin (Promega). Tryptic peptides were labeled using the Tandem Mass Tag (TMT) 6-plex Reagents (Thermo Fisher Scientific), pooled, and concentrated to 20 μl by vacuum centrifugation. The TMT-labeled peptides were then analyzed using a high-resolution nano-liquid chromatography with tandem mass spectrometry (LC–MS/MS) Tribrid system that included an Orbitrap Fusion with a Lumos coupled to an UltiMate 3,000 HPLC system (Thermo Fisher Scientific). Peptides (500 ng) were run on a pre-column (Acclaim PepMap 100, 75 μm × 2 cm, nanoViper, Thermo Fisher Scientific) and then on an analytical column (Acclaim PepMap RSCL, 75 μm × 50 cm, nanoViper, Thermo Fisher Scientific). The samples were eluted using a 90-min linear gradient of acetonitrile (4%–99%) in 0.1% formic acid.

All LC–MS/MS samples were analyzed using the Protein Discoverer software, version 2.1 (Thermo Fisher Scientific). Sequest HT was set up to search the Swiss-Prot database for both reviewed and unreviewed F. tularensis entries. The parameters for Sequest HT were set as follows: enzyme was trypsin, maximum missed cleavage was 2, precursor mass tolerance was 10 ppm, peptide tolerance was ±0.6 Da, fixed modifications were carbamidomethyl (C) and TMT sixplex (any N-terminus), and dynamic modifications was oxidation (M) and TMT sixplex (K). The parameters for Reporter Ion Quantifier were set as follows: integration tolerance was 20 ppm, integration method was most confident centroid, mass analyzer was FTMS, MS order was MS3, activation type was HCD, minimum collision energy was 0, and maximum collision energy was 1,000. The parameters for Percolator were set as follows: target FDR (strict) was 0.01, target FDR (relaxed) was 0.05, and validation was based on q-value.

The Proteome Discoverer software version 2.1 (Thermo Fisher Scientific) was again used to normalize the total peptide count for each sample. The algorithm summarizes the peptide group abundance for each sample and determines the maximum sum for all files to obtain a normalization factor. After normalization, the Reporter Ion Quantifier node performs scaling so that the average of all channels is obtained, and the node scales the abundance values of each sample so that the average of all control samples is 100. All other samples are then scaled up or down relative to 100 and when using multiplexed files, the node processes the samples from each file separately. Protein samples were prepared for each of the four F. tularensis strains in two independent experiments.

To determine if differential gene expression was significant, three statistical tests were applied amongst the four F. tularensis strains, specifically the two-sample T-test, the Mann–Whitney U-test, and the two-sample Kolmogorov–Smirnov test. The 24 sets of RNA-seq data for each gene in each strain was compared pairwise to each of the other RNA-Seq sets for the same gene in other strains, using the above three statistical tests. A differentially expressed gene was defined as significant if the value of p value of all statistical tests for all the pairwise comparisons was less than 0.01 (value of p < 0.01).

For RT-qPCR analyses, two-way ANOVA followed by Tukey’s post hoc tests were used for statistical analysis when multiple groups were analyzed. Data are presented as means ± SEM and are representative of triplicates in not less than three independent experiments. GraphPad Prism was used for the statistical analyses.

To characterize potential differences in the expression of metabolic enzymes within the F. tularensis subpopulations, deep RNA sequencing was performed for all three clades. RNA-Seq was performed on representative select agent F. tularensis strains that included hypervirulent subtype A.I strains SCHU S4 (A.I prototype) and MA00-2987 and virulent subtype A.II strain WY96-3418 (A.II prototype). Attenuated type B LVS was also evaluated since this non-select agent strain is often studied as a surrogate for this species and may serve as a reference. These strains were cultivated to mid-log in BHI, since growth in this medium recapitulates gene expression by this intracellular pathogen during a macrophage infection (Hazlett et al., 2008). Analysis of the 24 RNA-Seq libraries for each of the four F. tularensis strains revealed that most of the genes within the speHEA and aguAB operons were transcribed at higher levels in the hypervirulent A.I strains compared to subtype A.II WY96-3418 and attenuated type B LVS (Figure 1A). Examination of the genomic organization and content of speHEA and aguAB in the F. tularensis A.I, A.II, and B clades revealed that speHEA and aguAB were encoded in two adjacent operons (Figure 1A; Supplementary Figure S1). Further, only the A.I genomes contained intact speHEA and aguAB operons in which all the genes were full-length (Figure 1A; Supplementary Figure S1). More specifically, in the A.II and B genomes, speA was truncated. In type B strains, aguA and aguB were not full-length genes; aguA was disrupted by the insertion sequence element ISFtu1 and aguB had a premature stop codon (Figure 1A; Supplementary Figure S1). Together these analyses revealed that the genomic content of speHEA/aguAB differed between the F. tularensis A.I, A.II, and B clades, but was conserved within each subpopulation, and that only the A.I strains contained all intact genes within these operons.

To verify that the speHEA and aguAB genes were differentially transcribed in the A.I strains SCHU S4 and MA00-2987, A.II WY96-3418, and type B LVS, as was initially shown in the deep RNA sequencing data, RT-qPCR was performed. These results showed that the F. tularensis speH transcript, which is the first gene in the spe operon was expressed at a 7-fold higher level in the F. tularensis A.I and A.II strains relative to this gene in type B LVS (Figures 1A,B). The adjacent speE gene was expressed at a 4- and 2.8-fold higher level in the F. tularensis A.I and A.II strains, respectively, compared to type B LVS. The transcripts for aguA and aguB were 2.7- and 3.2-fold higher in abundance in the A.I strains, respectively, compared to the A.II strain, whereas both aguA and aguB are disrupted in all type B strains including LVS. Collectively, these data confirmed the differential expression of F. tularensis speHEA and aguAB and a variable trait between the A.I, A.II, and B subpopulations.

In the F. tularensis A.I strains, the stop codon and start codons for speH and speE overlap by a single bp, speE and speA are separated by 25 bp, and aguA and aguB are separated by a single bp, whereas 218 bp separate speA and aguA. Canonical −35 and −10 promoter elements were only identified upstream of speH and aguA. More importantly, our RT-qPCR results confirmed that speHEA and aguAB are two separate operons, each with a separate promoter (Figure 1C).

An examination of the predicted function of F. tularensis speHEA and aguAB revealed that these genes, along with metK that is located at a different chromosomal location, encode enzymes involved in methionine and arginine catabolism and subsequent polyamine biosynthesis (Figure 2). However, unlike the speHEA and aguAB genes, metK transcript levels were similar and only slightly higher in the F. tularensis A.I SCHU S4 and MA00-2987strains relative to A.II WY96-3418 and type B LVS (Figure 1B). The genes within the speHEA and aguAB operons, along with metK in the A.I strains encode enzymes that contribute to the catabolism of arginine and methionine and the subsequent biosynthesis of the polyamines agmatine, putrescine, and spermidine (Figure 2). The metK gene product S-adenosylmethionine synthetase and ATP are required for the first step in converting methionine to S-adenosylmethionine (SAM; Figure 2). SAM is the principal methyl donor in transmethylation and an aminopropyl donor in polyamine synthesis (Chiang et al., 1996). The next metabolic step for the conversion of SAM to S-adenosylmethioninamine in F. tularensis requires the enzyme S-adenosylmethionine decarboxylase, which is encoded by speH (Figure 2). This process releases carbon dioxide while producing S-adenosylmethioninamine, one of two substrates required by spermidine synthase SpeE for spermidine biosynthesis in F. tularensis (Figure 2). The methionine catabolic enzymes MetK, SpeH, and SpeE are all full-length in the F. tularensis A.I, A.II, and B clades, in contrast to the arginine catabolic enzymes SpeA in both the A.II and B subpopulations and AguA and AguB in the type B clade (Figures 1, 2).

In F. tularensis, arginine decarboxylase SpeA is responsible for the catabolism of arginine to agmatine and is only full-length in the A.I strains due to premature stop codons within this gene in the A.II and B clades (Figures 1, 2). This catabolic step releases carbon dioxide while producing the substrate agmatine. Next, agmatine deiminase AguA catabolizes agmatine to produce N-carbamoylputrescine in both the A.I and A.II strains, but not the type B strains since this gene is disrupted by the IS element ISFtu1 (Figures 1, 2). N-carbamoylputrescine amidase AguB in the F. tularensis type A strains has dual enzymatic activity that can catabolizes both N-carbamoylputrescine and citrulline, producing putrescine and ornithine, respectively (Figure 2). The catabolism of these two substrates by AguB results in the production of carbon dioxide and ammonium in both metabolic reactions.

A comparison of the primary sequence for these arginine and methionine metabolic enzymes in the F. tularensis A.II and B clades relative to the associated enzyme in the A.I subpopulation was next assessed. SpeH had one conserved residue difference in which the A.I and B strains had a methionine at residue 19 and the A.II strains had a valine at this position (M19V). SpeE had two conserved residues (I186L in both A.II and B strains and K266R in A.II strains), one non-conserved amino acid in A.II strains (G189D), and one non-conserved residue in the type B clade (S107F), relative to this enzyme in A.I subpopulation. Since SpeA was only produced by the A.I clade and not the A.II and B subpopulations, no comparisons could be made for this enzyme. A comparison of AguA and AguB in the A.I and A.II strains revealed that AguA in A.II strains had three non-conservative residue differences (K62N, R199H, and Q253K), whereas AguB was identical in the two type A subpopulations. Relative to S-adenosylmethionine synthetase MetK in the A.I strains, A.II strains had two non-conserved amino acids (E67K and G98D) and type B strains had one non-conserved residue (T105A). Together these results indicated that S-adenosylmethionine decarboxylase SpeH was the most conserved enzyme in all three F. tularensis subpopulations in these comparisons, suggesting an important role in SAM catabolism. SpeH also provides one of the two substrates required for spermidine biosynthesis in the A.I strains (Figure 2), indicating an additional vital role within this species.

The chromosomal location and directionality of the speHEA/aguAB operons and metK were compared in the F. tularensis A.I, A.II, and B strains. These results revealed that the spatial location and directionality of the spe/agu operons differed the most between the A.I and A.II strains and were the most similar in the A.I and B clades, even though only speH and speE are intact in the type B strains (Figure 3). In contrast, metK in the A.I and A.II clades was positioned in a similar chromosomal location and direction, and the type B metK differed in both aspects relative to the type A strains. However, in all three F. tularensis subpopulations, metK was positioned near the gene encoding the bacterial replication initiator protein DnaA, with the A.I metK being the closest to dnaA (Figure 3). Although all F. tularensis subpopulations share numerous features and have numerous IS elements, the type B strains have considerably more IS elements, which has ostensibly contributed to the higher abundance of pseudogenes in this clade (Table 1). Moreover, the most abundant IS element, specifically ISFtu1 from the IS630 family of transposases was responsible for the disruption of aguA in the type B clade.

The genes adjacent to the speHEA/aguAB operons in the F. tularensis A.I, A.II, and B clades were conserved in organization and directionality, despite the genomic, transcriptomic, and proteomic differences within the spe/agu operons. In all three subpopulations, threonine synthase thrC1 was adjacent and 5′ to speH, while UDP-2,3-diacylglucosamine hydrolase lpxH was adjacent and 3′ to the agu operon (Table 2; Supplementary Figure S2A). The proteins encoding by thrC1 and lpxH both provide critical cellular functions; ThrC1 catalyzes the production of threonine and LpxH is involved in lipid A biosynthesis. The thrC1 gene was transcribed in the same direction as the speHEA/aguAB operons, whereas lpxH was expressed in the opposite orientation. A remnant of a beta-galactosidase gene encoding only the first 76 amino acids of the 656-residue full-length enzyme also separated thrC1 and speH and was transcribed in the opposite direction of these two genes. This beta-galactosidase gene remnant shared the highest homology to associated enzyme in Burkholderia and Neisseria.

The genes flanking metK were also highly conserved within the F. tularensis A.I, A.II, and B subpopulations. A gene encoding a fatty acid desaturase was located 5′ to metK, but in the opposite orientation, while rpsP was positioned 3′ to metK and also oriented in the opposing direction, indicating that metK is monocistronic (Table 2; Supplementary Figure S2B). The rpsP gene encodes the 30S ribosomal protein S16 that is highly expressed and required for protein translation. The conservation of the genes and associated regulatory regions adjacent to the speHEA/aguAB operons and metK gene in the different F. tularensis subpopulations may provide a fitness advantage to this facultative intracellular pathogen.

The F. tularensis promoter region for the speHEA operon was 586 bp in length in the A.I and A.II clades and 564 bp in the B strains due to a 22 bp deletion. The same 22 bp sequence was located directly after the threonine synthase thrC coding sequence in the type A strains, whereas in the type B clade this sequence was incorporated into the 3′ end of the slightly shorter thrC open reading frame. The type B strains also had a single adenine insertion and a SNP that was a thymine instead of a guanine compared to the type A strains in the spe promoter region. A putative OmpR-like response regulator binding sequence (5′-TTGTAGCA-3′) was located between the −35 and −10 region in the spe promoter and was conserved within the A.I, A.II, and B clades. Although type A strains encode two OmpR-like orphan response regulators that are 228 and 229 residues in length with 54% amino acid similarity, type B strains only encode one of these regulatory factors. These and other differences in regulatory capabilities between these subpopulations may have contributed to the reduced expression levels of speH and speE in type B LVS relative to the A.I and A.II strains (Figure 1B).

A comparison of the nucleotide content in the 218 bp region separating the spe and agu operons in the A.I, A.II, and B clades revealed four SNPs. These SNPs included three guanines and one thymine in subtype A.I; two guanines and two thymines in subtype A.II; and one guanine, two adenines, and one thymine in type B strains. One SNP comprised the last 3′ nucleotide in the Pribnow box/−10 region, one SNP followed the Shine-Dalgarno/ribosomal binding site, and the remaining SNPs were in the intergenic region between the spe and agu operons. Overall, the A.II and B clades have more A/T enrichment in the intergenic nucleotides between the spe and agu operons than the A.I strains.

The F. tularensis metK promoter region was 293 bp and highly conserved in the A.I, A.II, and B subpopulations with only two SNPs, neither of which were located in the predicted Pribnow box/−10 region, −35 region, nor Shine-Dalgarno/ribosomal binding site. One SNP was a thymine in the A.I and B clades and a cytosine in the A.II strains. The other SNP was an adenine in the A.I clade, whereas the A.II and B strains had a guanine in this position. Interestingly, the promoter for metK and the adjacent rpsP gene, which encodes the 16S ribosomal protein, share the same predicted Pribnow box/−10 region that was located approximately in the middle of this promoter region. In addition to rpsP, other essential genes that are highly expressed during bacterial replication were located nearby to metK, including genes that encode 50S ribosomal proteins, tRNAs, and DNA-directed RNA polymerase subunits. These analyses support the premise that metK and rpsP are co-regulated and transcribed during favorable growth conditions for rapid replication.

A global RNA-Seq analysis of the immortalized mouse monocyte cell line P388D1 infected with F. tularensis LVS reported a 6.2- and 4.8-fold increase in speH and speE at 4 h post infection (hpi), respectively, and then a decrease in their abundance by 8 hpi (Bent et al., 2013). These results suggest that speH and speE contribute to the phagosomal escape of this intracellular pathogen, even though speA, aguA, and aguB are not intact in type B strains. Another notable global transcriptome study showed that the expression of the polyamine synthesis genes speHEA, aguAB, and metK were upregulated in primary mouse macrophages infected with F. tularensis SCHU S4 (Wehrly et al., 2009). Collectively, these data demonstrate that these metabolic gene products provide an important role in the persistence of F. tularensis, particularly for hypervirulent A.I SCHU S4.

Untargeted proteomic data from subtype A.I strains SCHU S4 and MA00-2987, subtype A.II strain WY96-3418, and type B LVS grown in BHI to mid-log growth phase were obtained using TMT mass spectrometry. These results revealed that AguB was 3.7- and 4-fold more abundant in the A.I strains SCHU S4 and MA00-2987, respectively, relative to this enzyme in A.II WY96-3418. MetK levels were similar in the A.I strains SCHU S4 and MA00-2987 and were only 1.7- and 1.8-fold higher when compared to the abundance of this enzyme in the A.II and B strains, respectively. In SCHU S4 and MA00-2987, MetK was 2-fold more abundant than AguB. SpeH, SpeE, SpeA, and AguA, however, were not detected in these analyses.

A comparison of proteins produced by F. tularensis LVS grown in several different culture broths showed similar MetK levels, while SpeE abundance was low or undetectable (Holland et al., 2017). In another global proteomic investigation, MetK was detected in both LVS and SCHU S4, whereas AguA, AguB, SpeE were only identified in SCHU S4 with just one of the two biological replicates detecting SpeE (Lenco et al., 2009). Collectively, these results indicated that although all the F. tularensis clades can metabolize methionine to produce SAM and S-adenosylmethioninamine, only the hypervirulent A.I strains can produce the polyamines agmatine, putrescine, and spermidine de novo via the enzymes encoded by speHEA, aguAB, and metK.

To determine the contribution of spermidine synthase SpeE in F. tularensis, chromosomal speE deletion mutants were generated in hypervirulent A.I SCHU S4 and type B LVS, since this enzyme was intact in both clades. The methods used to generate markerless and in-frame speE deletion mutants (ΔspeE) in F. tularensis were carried out as previously described (Horzempa et al., 2010). PCR amplification of the speHEA chromosomal locus in the F. tularensis SCHU S4 ΔspeE and LVS ΔspeE mutants confirmed that the 0.8 kbp speE gene was deleted from the genome of the respective wild-type strain (Figure 4; Supplementary Figure S3). DNA sequencing of this region confirmed the expected content and verified that the adjacent genes speH and speA were not disrupted. RT-qPCR validated that the absence of speE mRNA in the F. tularensis ΔspeE mutants and that no polar effects occurred in the transcription of the adjacent upstream and downstream genes in the SCHU S4 ΔspeE mutant relative to wild-type SCHU S4 (Supplementary Figure S4).

To assess if the deletion of speE altered growth, a comparison of wild-type SCHU S4 and wild-type LVS to the associated and isogenic ΔspeE mutant was evaluated. These results showed that SCHU S4 ΔspeE grew slower than wild-type SCHU S4, whereas the LVS ΔspeE mutant grew faster than wild-type LVS in Chamberlain’s CDM, suggesting that spermidine synthase SpeE provides a fitness advantage to SCHU S4 and not LVS. To ensure that no polar and off-target effects occurred during the process of producing the SCHU S4 ΔspeE mutant that contributed to this phenotype, trans-complementation was performed. For these assessments, the SCHU S4 ΔspeE mutant was complemented in trans with the Francisella expression plasmid pFNLTP containing the constitutive groESL promoter upstream of speE (pFN/Prˍgro/speE). However, these complementation experiments were unsuccessful, so the native spe promoter was then utilized to regulate speE expression. The usage of the native spe promoter in pFNLTP (pFN/Prˍspe/speE) successfully complemented the SCHU S4 ΔspeE mutant in trans, resulting in growth comparable to wild-type SCHU S4 (Figure 5). As shown in the Figure 5 insert, speE transcripts were present in wild-type SCHU S4 with empty plasmid and the complemented SCHU S4ΔspeE mutant containing the pFN/Prˍspe/speE expression plasmid, but not the SCHU S4ΔspeE mutant with empty plasmid. These data validated that speE is not transcribed by F. tularensis SCHU S4 ΔspeE and that the reduced growth rate of this mutant can be complemented.

To determine if arginine and methionine concentrations contributed to the regulation of the transcriptional expression of the speHEA/aguAB operons and metK by F. tularensis SCHU S4 and the isogenic ΔspeE mutant, RT-qPCR was performed. For these experiments, the F. tularensis strains were cultured in modified CDM containing 0, 0.3, 3, or 6 mM of arginine or methionine. When the bacteria were in mid-log growth phase, RNA was isolated and prepared for RT-qPCR, as described in the Materials and Methods section. No significant change in transcript abundance was observed for the spe, agu, and metK genes (data not shown). Further, an examination of the transcript levels for the proposed arginine and methionine importers ArgP and MetNIQ, respectively, did not change during growth with different concentrations of these essential amino acids (data not shown). These results indicated that the transcriptional expression of F. tularensis A.I speHEA/aguAB operons and metK, as well as the amino acid transporters ArgP and MetNIQ is not regulated by just the levels of these essential amino acids.

Next, we sought to determine if stress conditions altered the transcriptional expression of the genes in the polyamine biosynthesis pathway in F. tularensis SCHU S4 and the isogenic ΔspeE mutant when exposed to acidic conditions, hydrogen peroxide, and nitric oxide. After exposure to a low pH of 5.2 in CDM adjusted with hydrochloric acid for 1 h, 1 mM hydrogen peroxide for 1 h, and 100 μM of the nitric oxide-generating compound S-nitroso-N-acetylpenicillamine (SNAP) for 0.5 h, RNA was extracted and evaluated by RT-qPCR. Again, these conditions did not significantly increase nor decrease the transcript levels for these genes (data not shown). Therefore, a combination of stress factors and/or different conditions may be needed to modify the transcriptional expression of the speHEA, aguAB, and metK genes in F. tularensis SCHU S4 and the isogenic ΔspeE mutant. Alternatively, these gene products may be regulated at the translational or post-translational level to promptly respond to internal and external nutritional status and stress.

Charity and associates showed that LVS speH transcript levels were 3.3- and 3.9-fold lower in ΔmglA and ΔsspA deletion mutants, respectively (Charity et al., 2007). Since MglA and SspA facilitate sigma factor RpoD (σ⁷⁰) binding to DNA to regulate virulence and virulence-enhancing genes (Travis et al., 2021), these findings suggest that these transcription factors regulate spe operon expression and that these genes may contribute to the virulence of F. tularensis. Therefore, our ongoing studies will investigate the various factors that regulate the polyamine biosynthesis genes speHEA, aguAB, and metK, along with the role of the associated metabolic enzymes in F. tularensis pathogenicity.

To evaluate the growth rate and yield of F. tularensis SCHU S4 and LVS relative to the isogenic ΔspeE mutants in different media, these strains were grown in Chamberlain’s CDM and in BHI. These assessments demonstrated that all four strains grew to a higher cell density in Chamberlain’s CDM, which contains the polyamine spermidine or spermine (Chamberlain, 1965), in comparison to BHI (Figure 6). These results validated that the components within Chamberlain’s CDM provide the appropriate provisions for more efficient in vitro growth by F. tularensis, as well as corroborates with the data obtained by others (Mc Gann et al., 2010).

In BHI, F. tularensis SCHU S4 and the SCHU S4 ΔspeE mutant differed in growth rate and maximum cell density (Figure 6A), whereas the growth of wild-type LVS and the LVS ΔspeE mutant was similar (Figure 6B). Unlike wild-type SCHU S4, the SCHU S4 ΔspeE mutant exhibited diauxic growth in BHI and grew at a faster rate than wild-type SCHU S4 during these two exponential phases. The SCHU S4 ΔspeE mutant also reached a slightly higher cell density (maximum OD₆₀₀ of 1.2) than wild-type SCHU S4 (maximum OD₆₀₀ of 1.0), which occurred after 25 h of growth in BHI (Figure 6A). However, after SCHU S4 ΔspeE reached maximum cell density, cell death occurred with a concomitant and continuous decrease in cell density. Unlike the SCHU S4 ΔspeE mutant, wild-type SCHU S4 did not grow in a diauxic manner in BHI, but rather exhibited the canonical bacterial lag, exponential growth, and stationary phases. After 26 h of growth in BHI, SCHU S4 cell density reached a maximum OD₆₀₀ of 1.0 and remained mainly unchanged in stationary phase for at least 48 h when the experiments were terminated. Total cell density based on the area under the growth curve for SCHU S4 and SCHU S4 ΔspeE grown in BHI was 32.13 ± 0.26 and 33.89 ± 0.20, respectively.

For both wild-type LVS and LVS ΔspeE cultured in BHI, the growth rate slightly decreased after 13 h until maximum cell density was obtained (OD₆₀₀ of 0.8), which occurred after a total of 22 h when the cells were entering stationary phase (Figure 6B). The overall cell density based on the area under the growth curve for LVS and LVS ΔspeE grown in BHI was 26.02 ± 0.22 and 25.28 ± 0.20, respectively. Since growth rate and cell density were similar between wild-type LVS and the isogenic ΔspeE mutant, these results indicated that SpeE spermidine synthase does not substantially contribute to the fitness of these type B strains when cultured in BHI. In contrast, these findings indicated that spermidine synthase SpeE promotes steady state growth in SCHU S4 and implicates that spermidine production by this enzyme may also provide a protective function against factors that promote cell death during growth in BHI.

In Chamberlain’s CDM, both F. tularensis SCHU S4 and SCHU S4 ΔspeE grew substantially faster than LVS and LVS ΔspeE, with wild-type SCHU S4 growing 3.8-fold faster than wild-type LVS. The generation time of 1.22 h was obtained for SCHU S4, whereas LVS had a doubling time of 4.66 h (Figure 6B). However, wild-type LVS and the LVS ΔspeE mutant eventually reached a similar maximum cell density after 33 h (OD₆₀₀ of 1.4) and 27 h (OD₆₀₀ of 1.53), respectively (Figure 6B). In comparison, wild-type SCHU S4 and the SCHU S4 ΔspeE mutant grown in CDM reached the highest cell density after 12 h (OD₆₀₀ of 1.6) and 17 h (OD₆₀₀ of 1.54), respectively (Figure 6A). Further, in CDM both LVS and the LVS ΔspeE mutant showed diauxic growth (Figure 6B), which was not the case for SCHU S4 and the SCHU S4 ΔspeE mutant (Figure 6A). In CDM, the total cell density based on the area under the growth curves for SCHU S4 and SCHU S4 ΔspeE was 50.07 ± 0.20 and 44.98 ± 0.12, respectively. For LVS and LVS ΔspeE grown in CDM, the overall cell density based on the area under the growth curves was 31.29 ± 0.33 and 36.61 ± 0.17, respectively. In general, these results demonstrated that spermidine synthase SpeE in wild-type SCHU S4 contributes to an enhanced growth rate and cell yield in CDM, whereas the deletion of speE in LVS augmented the growth rate and slightly increased cell yield in CDM (Figure 6).