Introduction

Polyamines are flexible aliphatic cations that are found in virtually all organisms. E. coli contains putrescine (1,4-diaminobutane) and lesser amounts of spermidine (N-(3-aminopropyl)-1,4-diaminobutane), and cadaverine (1,5-diaminopentane) (1, 2). Both the pathways and enzymes of polyamine synthesis are redundant (Fig. 1). Strains that lack eight of these genes do not contain detectable polyamines, grow slowly aerobically, and do not grow anaerobically (3). Polyamines interact with nucleic acids and phospholipids, can affect chromosome and ribosome structure, ribosome-mRNA interactions, and protein and nucleic acid elongation rates (4). Polyamines modulate, but are not required for, protein synthesis (5) and expression of hundreds of genes (6). A major mechanism of polyamine-dependent regulation is translational control of genes for 18 major transcription regulators: four of the seven σ factors – σ¹⁸ (FecI), σ²⁴ (RpoE), σ³⁸ (RpoS), and σ⁵⁴ (RpoN); three histone-like DNA-binding proteins (Fis, H-NS, and StpA); and CpxR, Cya, Cra, and UvrY (5, 6).

Figure 1

Polyamines control flagella-dependent surface motility of Proteus mirabilis and pili-dependent surface motility (PDSM) of E. coli: the latter has been reported to require the polyamine spermidine (7, 8). While reconstructing putrescine-deficient strains to characterize the basis for the anaerobic growth defect, we noticed that a mutant unable to synthesize spermidine moved normally, while mutants lacking enzymes of putrescine synthesis moved poorly. We characterized the role of polyamines on E. coli surface motility and pili synthesis by genetic analysis, ELISA assays, RT-qPCR, electron microscopy, and RNAseq. We showed that PDSM and pili synthesis required two independent pathways of putrescine synthesis, but did not require spermidine, and was modulated by putrescine transport and catabolism. In addition to confirming that putrescine controlled pili synthesis, results from an RNAseq analysis identified regulators that mediated putrescine-dependent control, and, unexpectedly, suggested a putrescine homeostatic network that rewires metabolism to maintain intracellular putrescine.

Results

Pili-dependent surface motility (PDSM) of W3110

While characterizing mutants lacking putrescine synthetic enzymes, we found that some strains were defective in surface motility. Because any single laboratory strain may have properties that result from unintended laboratory evolution, we assessed nine strains for their surface motility pattern. Five strains covered a plate in 12-18 hours, while four did not traverse a plate in 36 hours. The slow-moving strains generated genetically stable fast-moving variants which suggests that the former were ancestral. Our lab strain of W3110 moved slowly, and among the slow movers showed the least variability in spreading diameter and generated fewer fast-moving variants. Its outward movement formed concentric rings, with alternating phases of movement and consolidation/growth (Fig 2A). Deletion of fimA, which codes for the major type 1 pili subunit, abolished the oscillatory pattern and reduced, but did not eliminate, surface motility, but deletion of fliC which encodes the major flagellum subunit did not affect surface motility (Fig. 2A). Electron microscopy confirmed that W3110-LR isolated from surface motility plates possessed pili (Fig. 2B). None of >500 cells expressed flagella. A mixed population of elongated (3-4 µm) and non-elongated (<2 µm) cells were visible, and pili were mostly associated with elongated cells. Furthermore, W3110 surface motility requires glucose (9) which should prevent flagella synthesis via inhibition of cyclic AMP synthesis (10). We confirmed that glucose controlled flagella synthesis in W3110 by its suppression of swimming motility (Fig 2C). A W3110-LR derivative with deletions of both fliC and fimA still exhibited outward movement on a surface motility plate (Fig 2A), but electron microscopy showed the absence of appendages (Fig. 2C). In summary, genetics, electron microscopy, and a glucose requirement show that W3110 surface motility requires pili.

Figure 2.

PDSM required two independent putrescine synthesis pathways

We examined PDSM requirements for polyamine anabolic pathways. A major putrescine biosynthetic pathway is ornithine decarboxylation either by the constitutive SpeC or the low-pH inducible SpeF. The ΔspeC and ΔspeF mutants moved as well as the parental strain, but a ΔspeC ΔspeF double mutant moved less well (Figure 3). SpeA (arginine decarboxylase) and SpeB (agmatinase) catalyze an alternate two-step putrescine synthetic pathway (Fig 1) (11): deletion of either gene reduced surface motility (Figure 3). The DspeA, DspeB, and DspeC DspeF strains grew as well as the parental strain (Supplemental Figure 1A) and had normal swimming motility (Supplemental Figure 1B). A ΔspeE mutant, which lacks the enzyme for spermidine synthesis, moved on a surface as well as the parental strain (Fig 3).

Figure 3.

Supplemental putrescine at 1 mM restored surface motility of a speB mutant (Figure 4A) and a speA mutant (not shown). Spermidine did not restore wild type movement and, also, caused an unusual pattern of movement in the parental strain (Figure 4B). These exogenous concentrations of putrescine and spermidine had no effect on growth rates in liquid motility medium (not shown). In summary, PDSM required the simultaneous presence of two independent pathways of putrescine synthesis but did not require spermidine.

Figure 4:

Loss of putrescine transport systems affected surface motility

Putrescine transport may maintain an appropriate intracellular putrescine concentration, because E. coli secretes putrescine (12). E. coli has several putrescine transport systems (13, 14). Mutants lacking the PlaP and PotF putrescine transporters had 50% and 40% reduced surface motility, respectively, and lost the concentric ring pattern (Figure 5A). Loss of the PotE and PuuP transporters affected neither the movement diameter (Figure 5B) nor concentric ring formation (Figure 5A). We note that potE and plaP are more highly expressed than potE and puuP (85, 64, 17, and 6 counts per million transcripts, respectively) for W3110 grown in liquid motility medium (15). We conclude that extracellular putrescine and its transport contribute to PDSM.

Figure 5:

Putrescine catabolism affected surface motility

Because putrescine catabolism modulates intracellular putrescine concentrations (16, 17), we examined PDSM in putrescine catabolic mutants. The major putrescine catabolic pathway is initiated with putrescine glutamylation (reactions 6-9, and 4-5, Fig 6A), and a second pathway initiates with putrescine deamination (reactions 1-5, Fig 6A) (17). Deletion of either puuA or patA which code for the first enzymes of their respective pathways did not affect surface motility (Fig 6B). A double mutant moved less well which was unexpected because the proposed higher intracellular putrescine should have stimulated PDSM. Consistent with the stimulatory effect of putrescine, loss of puuR which codes for the repressor of PuuA-initiated pathway genes (17), impaired PDSM (Fig 6B). One possible explanation to reconcile these seemingly contradictory results is that an optimal polyamine concentration stimulates PDSM, and a high concentration is inhibitory. Experiments that are described in the next section test this possibility.

Figure 6:

High putrescine reduced expression of pili genes

High putrescine inhibits translation of some mRNAs, and high spermidine, a product of putrescine metabolism, inhibits growth (18, 19). If the phenotype of the ΔpatA ΔpuuA double mutant results from spermidine toxicity, then loss of spermidine synthase (SpeE) should reverse the phenotype. This prediction was not met which argues against spermidine toxicity (Fig 6C). For the speB mutant, 1 mM putrescine stimulated PDSM, and 4 mM putrescine was slightly inhibitory but eliminated the ring pattern (Fig 7AB). The concentric ring pattern was observed with 1 mM, but not 4 mM putrescine. Indirect enzyme linked immunosorbent assays (ELISAs) against FimA showed that the level of FimA in the speB mutant at 1 mM putrescine was higher than with 4 mM putrescine (Fig. 7C). Variation of supplemental putrescine did not affect FimA in the parental strain (Fig. 7C).

Figure 7:

To test transcriptional regulation of pili expression by putrescine, we examined transcript levels by RT-qPCR. Optimal fimA transcription occurred at 1 mM putrescine in the speB mutant (Fig 7D). Transmission electron microscopy (TEM) showed that pili expression was highest for the speB mutant at 1 mM putrescine (Fig. 8), moderate at 0.1- and 4-mM putrescine, and undetectable without putrescine. The cells grown with 4 mM putrescine were shorter and thinner, which suggests a stress, possibly mimicking osmotic stress. From these results and those in the previous section, we conclude that pili expression required an optimal putrescine concentration.

Figure 8:

RNAseq analysis identified H-NS as a possible mediator of putrescine-dependent control

To determine whether putrescine affects transcription of genes whose products control fim operon expression (20), we performed RNAseq analysis of the parental and speB strains grown with and without 1 mM putrescine. The transcriptomes of the parental strain both with and without putrescine and the speB mutant with 1 mM putrescine were highly similar with R² values of 0.950 and 0.954, respectively, based on expression graphs comparing the transcriptomes of the parental strain in both conditions to the speB mutant strain (Supplemental Figure 2). Compared to the transcriptome of the speB mutant grown without putrescine, the transcriptomes of the parental strain without putrescine (R² = 0.796), the parental with 1 mM putrescine (R² = 0.786), and the speB mutant with 1 mM putrescine (R² = 0.820) differed substantially. A multidimensional scaling plot and a plot of the 100 most variable genes further visualize these interactions (Fig 9A and 9B).

Figure 9:

When compared to the parental strain, the speB mutant grown without putrescine had 310 downregulated genes and 159 upregulated genes with at least four-fold differential expression and FDR < 0.05. Transcripts for the putrescine-responsive puuAP and puuDRCBE operons, which specify genes of the major putrescine catabolic pathway, were reduced from 1.6-to 14-fold (FDR ≤ 0.02) in the speB mutant (Supplemental Table 1), which implies lower intracellular putrescine. Deletion of speB reduced transcripts for genes of the fimA operon and fimE, but not of fimB (Fig 10A). Furthermore, loss of speB affected some of the known positive effectors that control fim operon expression which include Fis, H-NS, IHF, Lrp, and QseB (21). Transcripts from hns were reduced, transcripts from ihfA and ihfB were not affected, and transcripts from fis, lrp, and qseB were increased when speB was deleted (Fig 10B). Loss of speB also increased transcripts from lrhA which codes for a positive effector of fimE. Loss of several of these regulators eliminated PDSM which was expected (Fig 10C). H-NS may mediate putrescine-dependent control of pili synthesis. However, given the numerous effects of loss of speB, a generalized dysregulation of several major transcription factors may also produce a less-than-optimal constellation of regulators.

Figure 10:

Evidence for a homeostatic network that maintains intracellular putrescine

The gene expression changes resulting from loss of speB unexpectedly suggested compensatory mechanisms for low putrescine. The transcript differences for all genes are provided in Supplemental Table 1, and for selected genes in Table 1. A gene expression pattern is summarized in Fig 11. The speB mutant had more transcripts from (a) genes of arginine and ornithine carboxylase, (b) all genes of ornithine and arginine synthesis from glutamate, and (c) genes for three separate putrescine transport systems. The speB mutant also had fewer transcripts from the sap operon which codes for a putrescine exporter (12), and from rpmH and rpsT, whose products inhibit ornithine and arginine decarboxylase (22). The net effect of these changes should be to increase intracellular putrescine: follow the solid arrows in the left half of Fig. 11.

Table 1.

Figure 11:

Loss of speB also affected pathways of energy metabolism. For instance, glutamate and α-ketoglutarate are at a branch point in metabolism which can either provide putrescine precursors or generate energy. The speB mutant had fewer transcripts for metabolic genes coding for all citric acid cycle enzymes, NADH dehydrogenase I (the entire nuo operon), cytochrome oxidase (the cyo operon), and several genes of menaquinone (the menFDHBCE operon), ubiquinone, and heme synthesis. In addition to these differences, the speB mutant also had fewer transcripts for all genes of iron acquisition, which is consistent with diminished energy generation. In sum, this pattern suggests that low putrescine diverts metabolic flux from energy-generating pathways to putrescine synthesis.

One of the most differentially expressed gene between the parental and speB mutant strains was mgtA which had 60-fold more transcripts in the speB mutant and codes for the major magnesium transporter. The speB mutant also had more transcripts for several phosphate assimilation genes, including phoB which codes for the response regulator of phosphate limitation.

Discussion

PDSM and pili synthesis required putrescine synthesis and were affected by putrescine degradation and transport, as well as several transcription factors, which implies that numerous factors control intracellular putrescine concentrations. The effect of loss of putrescine transport systems on PDSM implicates extracellular putrescine on putrescine homeostasis. A requirement for the SpeA-SpeB pathway implicates extracellular arginine since SpeA is mostly periplasmic (23). The role of both putrescine synthesis and degradation suggests rapid adjustments to fluctuating putrescine concentrations. Numerous transcription factors control pili synthesis. Of these, only the gene for H-NS had fewer transcripts in the speB mutant, which is consistent with H-NS mediating putrescine-dependent control. This conclusion is not considered strong because of the only two-fold transcript reduction, and because of numerous changes in transcripts for other potential pili regulations in the speB mutant, albeit in the wrong direction. Therefore, we cannot exclude a general regulatory dysfunction mediated by depleted levels of putrescine as the basis for loss of pili synthesis.

An unexpected observation was that the speB mutant had more transcripts for genes whose products will maintain intracellular putrescine, i.e., compensatory mechanisms for low intracellular putrescine (Figure 11). Conversely, loss of speB lowered transcripts for many genes of energy metabolism and all genes of iron acquisition. This gene expression pattern suggests a diversion of metabolism away from energy generation and toward putrescine synthesis, i.e., a prioritization of putrescine synthesis over aerobic energy metabolism.

We also propose that other changes are also compensatory mechanisms for low intracellular putrescine, specifically, more transcripts for genes coding for magnesium and phosphate transport systems in the speB mutant. To the extent that magnesium can directly replace putrescine, e.g., binding to nucleotides and RNA (24), magnesium could compensate for low putrescine. Our observations confirm those of J. Slauch and colleagues on an inverse correlation between intracellular magnesium and polyamines in Salmonella (25). They proposed an overall divalent cation homeostasis maintained by either magnesium or the polyamines (25). A rationale for the effect on phosphate metabolism is not obvious but considering the effects of low putrescine on energy metabolism, increased phosphate transport may be an adjustment to changes in energy metabolism. In any case, the substantial changes in gene expression due to low putrescine suggest multiple compensatory mechanisms, and a homeostatic gene network which because of effects on metabolic genes can rewire metabolism.

In contrast to our results, a thorough and well-designed study showed that E. coli PDSM requires spermidine instead of putrescine (8). The differences between the studies were strain backgrounds, media (glucose-LB vs glucose-tryptone), and incubation temperature (33⁰ vs 37⁰). We found PDSM results for 37⁰ incubations were highly variable and resulted in numerous fast-moving variants. Since growth at the higher temperature is faster, and faster growth is associated with higher putrescine, we propose that growth at 37⁰ results in an inhibitory putrescine concentration. In this case, spermidine would be stimulatory because of its known inhibition of the putrescine biosynthetic enzymes SpeA and SpeC (26, 27) and subsequent reduction of intracellular putrescine to an optimal concentration.

The transcript changes observed in our study were different from those in the only other systematic study of putrescine’s effect on E. coli gene expression (6). Both studies had almost identical experimental design: 1.0- or 1.14-mM putrescine was added to either a speB or speB speC mutant, respectively. Differences in strains, growth media (minimal media vs tryptone-containing media), and growth temperature (33⁰ vs 37⁰) could account for the discrepancy in results. The most crucial difference could be the growth media: Igarashi and colleagues grew bacteria in a minimal medium, while we grew bacteria in an amino acid-containing medium that was required for surface motility. We propose that the slower growth in minimal medium correlates with higher guanosine tetraphosphate and in turn counters and obscures many growth-stimulatory putrescine effects. In contrast, the growth stimulation from the amino acids in our medium would not counter putrescine effects and would essentially allow detection of effects caused solely by changes in putrescine concentrations. Regardless of the explanation, common results can be interpreted as core responses to putrescine limitation. Both studies observed that low putrescine increased transcripts for genes of arginine synthesis and magnesium transport, and reduced transcripts for the sdhCDAB-sucABCD operon which codes for TCA cycle enzymes. The link between putrescine and arginine synthesis is probably mediated by the ArgR repressor because argR transcripts are 2.5-fold lower (FDR = 0.0004) in the speB mutant (Supplemental Table 1). Cyclic AMP may mediate putrescine-dependent control of the sdh-suc operon because putrescine controls translation of cya mRNA, which codes for adenylate cyclase, and cyclic-AMP controls the sdhCDAB-sucABCD operon. As noted above, the link between putrescine and magnesium transport is a recent observation (25).

Several results implicate putrescine in E. coli virulence during urinary tract infections. First, pili-dependent binding to the urothelium is an essential step for uropathogenic E. coli (28). Second, putrescine is present in urine from infected, but not healthy individuals (29, 30). Third, urinary putrescine can result from UPEC growth in urine (30) and urothelial pathophysiology (31). Putrescine’s possible role in virulence is seemingly contradicted by the observation that a speB mutant outcompetes the parental CFT073 in the bladder of mice, i.e., less putrescine synthesis enhanced virulence (32). A majority of uropathogenic E. coli are associated with phylogenetic group B2―including CFT073, whereas most E. coli lab strains, such as W3110, are group A. Group B2 strains lack genes for the major putrescine catabolic pathway and have significantly higher transcripts for genes coding for SpeA and SpeB (15). Both observations suggest that the cytoplasmic putrescine concentration is higher in group B2 strains. We propose that the putrescine concentration in group B2 strains is normally in the inhibitory range for pili synthesis, and that loss of SpeB reduces putrescine to a level that stimulates pili synthesis. Regardless of the explanation, putrescine is likely to be important during urinary tract infections, but aspects of putrescine control of pili synthesis and other processes could be different in uropathogenic E. coli and is worth further examination.

Experimental Procedures

Strains

All strains used for growth rate determinations and motility assays were derivatives of E. coli K-12 strain W3110 and are listed in Table 2. Great variations exist in standard lab strains, including W3110, from different labs. We tested nine strains, most from the Coli Genetic Stock Center, and chose our lab strain because of relative genetic stability and more quantitatively reproducible results. Our strain originated from the lab of Jon Beckwith in the 1960s via the lab of Boris Magasanik where it had been stored in a room temperature stab until the mid-1980s, at which time it was revived and frozen at -80⁰ C. Although mutations may have accumulated during storage, its transcriptome is similar to that of other strains of the same phylogenetic group when grown in motility medium (15). To construct the mutant strains, the altered allele was obtained from the Keio collection in which the gene of interest had been deleted and replaced with an antibiotic resistance gene (33). The marked deletion allele was transferred into W3110 by P1 transduction (34). The antibiotic gene was removed as described which generated an in-frame deletion (35).

Table 1.

Media and Growth Conditions

For P1 transductions and plasmid transformation experiments, cells were grown in standard LB liquid medium (1% tryptone, 1% NaCl and 0.5% yeast extract, pH 7.0) at 37°C. Antibiotics were used for selection at concentrations of 25 µg/mL (both chloramphenicol and kanamycin). For growth analysis, bacteria were grown in liquid motility medium (0.5% glucose, 1% tryptone, and 0.25% NaCl) which we refer to as GT medium. Starter cultures (typically 6–12 h incubation) were grown in GT medium, harvested by centrifugation, washed twice with phosphate buffered saline and re-suspended in GT medium before inoculation. The cells were then grown at 37° C in aerobic conditions (shaking at 240 rpm) and the turbidity was measured every thirty minutes. Cell growth was measured in Klett units using a Scienceware Klett colorimeter with a KS-54 filter. 100 Klett units represent an OD₆₀₀ value of about 0.7.

Surface motility assay

For a standard surface motility assay, single colonies from fresh plates (streaked out from frozen stocks a day before) were inoculated in GT medium for six hours at 37° C in aerobic conditions (shaking at 240 rpm). 30 ml of autoclaved GT medium with 0.45% agar (Eiken, Tokyo, Japan) was poured into sterile polystyrene petri-dish (100 mm X 15 mm) and allowed to solidify at room temperature for approximately six hours. Then, 1 µL of the pre-motility growth medium was inoculated at the center of the agar plate and incubated at 33° C for 36 hrs. Each experiment was performed in triplicate and pictures were taken after 36 hours. Surface motility was extremely sensitive to humidity. Opening the incubator before 36 hours resulted in movement cessation. The surface motility assay was performed at 33⁰ C because results from incubations at 37⁰ C were highly variable and the cultures more frequently generated genetically stable fast-moving variants.

Swim assay

The media and culturing for swimming motility is identical to that for surface motility, except that the plates contained 0.25% agar and were solidified at room temperature for about one hour before inoculation. Then, 1 µL of the culture was stabbed in the center of this swim agar plate and incubated at 33° C for 20 hours. Each experiment was performed in triplicate.

Transmission electron microscopy

Cells from surface motility plates were collected and fixed with 2.5% glutaraldehyde. Bacteria were absorbed onto Foamvar carbon-coated copper grids for 1 min. Grids were washed with distilled water and stained with 1% phosphotungstic acid for 30 s. Washed and stained grids were dried at 37⁰C for 10 minutes. Samples were viewed on a JEOL 1200 EX transmission electron microscope at the University of Texas Southwestern (Figure 2) and the University of Texas at Dallas (Figure 8).

ELISA analysis

Cells were grown overnight in 5 mL GT media followed by a 2-hour growth in GT media with 0.0, 0.1, 1.0, or 4.0 mM putrescine. Cells were then lysed using a 24-gauge needle. The cell lysate was diluted to a concentration of 1 mg/mL protein (based on A280) in pH 9.6 coating buffer (3 g Na₂CO₃, 6 g NaHCO₃, 1000 mL distilled water) and then coated onto the walls of a 96-well plate by incubating overnight at 4°C. The wells were rinsed 3 times with PBS and further blocked using coating buffer with 1% bovine albumin overnight at 4°C. The following day, wells were rinsed 3 times with PBS, and primary antibodies to FimA or RpoD were added following supplier directions. After a 2-hour room temperature incubation, secondary antibodies conjugated to HRP were added and incubated for a further two hours. HRP was activated using the TMB substrate kit (Thermo-34021) following the provided protocol. Expression was read on a BioTek plate reader based on the provided kit protocol. Data was blanked to wells only treated with bovine albumin then normalized to RpoD. Data was generated from technical and biological triplicates.

DNA isolation and genome assembly and annotation for W3110

DNA was isolated based on previous established protocols (36). Short reads were sequenced at the SeqCenter (Pittsburgh, Pennsylvania) using Illumina tagmentation-based and PCR-based DNA prep and custom IDT indices targeting inserts of 280 bp without further fragmentation or size selection steps. The Illumina NovaSeq X Plus sequencer was ran producing 2X151 paired-end reads.

Demultiplexing, quality control, and adapter trimmer was performed with bcl-convert (v4.2.4). Total Illumina Reads (R1+R2): 4059078 with 553087045 bps >Q30.

Long reads were prepared using the PCR-free Oxford Nanopore Technologies ligation sequencing kit with the NEBNext companion module. No fragmentation or size selection was performed. Long read libraries were performed using R10.4.1 flow cell. The 400bps sequencing mode with a minimum read length of 200bp were selected. Gupply (v6.5.7) was used at the super-accurate basecalling, demultiplexing, and adapter removal. Total long reads: 53773 with 89.607 % of bps > Q20.

To generate the completed genome, porechop (v0.2.4) was used to trim residual adapter sequences from long reads. Flye (v2.9.2) was used to generate the de novo assembly under the nano-hq model. Reads longer than the estimated N₅₀ based on a genome size of 6Mbp initiated the assembly. Subsequent polishing using the short read data was performed using Pilon (v1.24) under default parameters. Long read contigs with an average short read coverage of 15X or less were removed from the assembly. The assembled contig was confirmed to be circular via circulator (v1.5.5). Annotation was performed using prokka (v 1.14.6). Finally, statistics were recorded using QUAST (v5.2.0). The final genome contained 1 contig of 4750347 bp with a sequencing depth of 123.92x. The N₅₀ was 4750347. This genome can be accessed via the accession number CP165600 or via the BioProject PRJNA1142534 via NCBI.

The completed genome has an estimated average nucleotide identity of 99.9694 to the W3110 genome deposited to NCBI (genome assembly ASM1024v1).

RNA isolation and quality control

Cells were grown for 2 hours in 1 mL GT media. 60 µL were then added to 1 mL of fresh GT, with or without 1 mM putrescine, and grown for another two hours. After growth, the cells were centrifuged, and frozen at -80⁰C. The isolation and analysis protocol has been previously published (15). Cell pellets were thawed, resuspended in 0.7 mL of buffer RLT (Qiagen RNeasy kit), and mechanically lysed using a bead beater (FastPrep-24 Classic from MP Biomedical) set at the highest setting for three 45-s cycles with a 5-min rest period on ice between cycles. Cell lysates were used to isolate RNA using a Qiagen RNeasy mini kit. RNA was quantified, DNase treated, and re-quantified on a nanodrop. RNA that passed the first quality check was analyzed on RNA-IQ (Qubit). RNA with an IQ higher than 7.5 was used. RNA that passed both checks were run on an agarose gel to check for RNA integrity. RT-qPCR was used to ensure there was no DNA contamination. RNA libraries that passed all these quality checks were submitted to the genomics core facility at the University of Texas at Dallas for RNA sequencing. The core performed rRNA removal (RiboMinus Transcriptome Isolation Kit or RiboCop bacterial rRNA depletion—Lexogen), library preparation (Stranded Total RNA Prep—Illumina), and single-end Illumina sequencing.

Reverse transcription quantitative PCR

RNA was isolated as described above. One microgram of RNA was reverse transcribed using LunaScript RT Super Mix (NEB E3010) and another microgram was subjected to the same reaction as the reverse transcribed RNA, except without reverse transcriptase (negative control) following manufacturer instructions. The cDNA was then diluted to 10 ng/µL for qPCR. PCR reactions were composed of 10 ng cDNA, 8 µL of nuclease free water, 10 nM primers (rpoD: Forward: TCGTGTTGAAGCAGAAGAAGCG; Reverse: TCGTCATCGCCATCTTCTTCG) (fimA: Forward: ATGGTGGGACCGTTCACTTT; Reverse: GGCAACAGCGGCTTTAGATG), and PowerUp SYBR Green 2X master mix for qPCR (ThermoFisher A25777) in a 20 µL reaction. A Quantstudio 3 qPCR machine was used to generate critical threshold (Ct) values. The Ct values were then analyzed using the 2^-ΔΔC_T method (37) using rpoD as a library control.

RNAseq Analysis

Transcripts were aligned to the W3110 genome and CDS (CP165600) available on NCBI using CLC genomics workbench (Qiagen version 22.00) for RNA-seq analysis. The gene expression counts generated as this output were used to perform differentially expressed gene (DEG) analysis using EdgeR (version 3.38.4) to analyze read counts (38, 39). DEGs had an adjusted P-value less than 0.05.

Data availability

The raw RNA seq reads were deposited in GenBank: BioProject accession number PRJNA1126736. The genome was deposited in GenBank: BioProject accession number PRJNA1142534. The genome accession number is CP165600.

Acknowledgements

This work was funded through private donations. PEZ received endowment support from the Cain Foundation through the Felecia and John Cain Distinguished Chair in Women’s Health in honor of Philippe E. Zimmern, MD. The authors declare no potential conflicts of interests.

The authors (JH, SH, and GV) would like to thank Prof. Kelli Palmer of the University of Texas at Dallas for the use of her space, supplies, and tissue homogenizer. JH would like to further thank Prof. Palmer for the use of her CLC genomics program. The authors would like to thank the Genome Center at The University of Texas at Dallas for the services to support our research. The authors would like to thank the Olympus Discovery Center/Imaging Core facility at UT Dallas for providing equipment and support (Fig. 8), and the UT Southwestern Medical Center for electron microscopy (Fig. 2).