Differential susceptibility of airway and ocular surface cell lines to FlhDC-mediated virulence factors PhlA and ShlA from Serratia marcescens

Abstract

Introduction

Serratia marcescens is a bacterial pathogen that causes ventilator-associated pneumonia and ocular infections. The FlhD and FlhC proteins complex to form a heteromeric transcription factor whose regulon, in S. marcescens , regulates genes for the production of flagellum, phospholipase A and the cytolysin ShlA. The previously identified mutation, scrp-31, resulted in highly elevated expression of the flhDC operon. The scrp-31 mutant was observed to be more cytotoxic to human airway and ocular surface epithelial cells than the wild-type bacteria and the present study sought to identify the mechanism underlying the increased cytotoxicity phenotype.

Hypothesis/Gap Statement

Although FlhC and FlhD have been implicated as virulence determinants, the mechanisms by which these proteins regulate bacterial cytotoxicity to different cell types remains unclear.

Aim

This study aimed to evaluate the mechanisms of FlhDC-mediated cytotoxicity to human epithelial cells by S. marcescens .

Methodology

Wild-type and mutant bacteria and bacterial secretomes were used to challenge airway and ocular surface cell lines as evaluated by resazurin and calcein AM staining. Pathogenesis was further tested using a Galleria mellonella infection model.

Results

The increased cytotoxicity of scrp-31 bacteria and secretomes to both cell lines was eliminated by mutation of flhD and shlA. Mutation of the flagellin gene had no impact on cytotoxicity under any tested condition. Elimination of the phospholipase gene, phlA, had no effect on bacteria-induced cytotoxicity to either cell line, but reduced cytotoxicity caused by secretomes to airway epithelial cells. Mutation of flhD and shlA, but not phlA, reduced bacterial killing of G. mellonella larvae.

Conclusion

This study indicates that the S. marcescens FlhDC-regulated secreted proteins PhlA and ShlA, but not flagellin, are cytotoxic to airway and ocular surface cells and demonstrates differences in human epithelial cell susceptibility to PhlA.

Introduction

FlhC and FlhD proteins form a heterodimer that is the master regulator of flagella production in numerous enteric bacteria, such as Escherichia coli , Proteus mirabilis , Salmonella typhimurium and Yersinia enterocolitica [1–3]. Evidence from several bacterial genera support the importance of FlhDC as a regulator of virulence. An flhDC deletion mutant of Edwardsiella tarda exhibited reduced virulence in a zebrafish infection model [4]. With Salmonella enterica serovar Typhi, mutation of flhDC produced bacteria with reduced ability to invade human intestinal epithelial Henle-407 cells and decreased ability to produce cytotoxicity to mouse macrophage-like cell line J774.A1 cells; these findings were correlated with a reduced expression of type III secretion system-related genes [5]. In a variety of bacteria, FlhDC has been shown to regulate a number of factors that probably play a role in pathogenesis. For example, FlhD and FhlC were reported to mediate biofilm and pellicle formation [6, 7], and expression of the phospholipase [8, 9], lipase, protease and haemolysin genes [1, 10]. Together, these and other studies support the suggestion that FlhDC is important in a broad range of gram-negative bacteria for control of motility and virulence determinants.

In the opportunistic pathogen Serratia marcescens , FlhC and FlhD were shown to positively regulate flagella and three other secreted proteins. These are the secreted nuclease [11], phospholipase A (PhlA) [12–17] and the surface-associated haemolysin/cytolysin, ShlA [16, 18]. Genetic evidence also suggests that flhDC negatively influences expression of type I pilus biosynthesis and associated biofilm formation by S. marcescens [19].

ShlA is a well-studied pore-forming toxin [20] that is important for cytotoxicity in a number of in vitro cell systems [18, 21–25], and for pathogenesis in invertebrate [25–27] and mammalian infection models [16, 28]. It has been implicated as a key determinant for invasion into mammalian cells [18, 29] and activation of autophagy by intracellular and extracellular S. marcescens [18, 30], and for escape of internalized S. marcescens from host cells [31].

Fedrigo et al. [32] showed that S. marcescens flhD mutants had reduced adherence to and invasion into Chinese hamster ovary cells, and reduced induction of autophagy. These effects were concluded to be flagellum-dependent and PhlA exoenzyme-independent. However, it could also be due to other FlhD2C2-regulated molecules, as the role of the flagellum was not directly evaluated. Further studies by the Véscovi group clarified that the cytolysin protein ShlA is responsible for induction of autophagy caused by internalized S. marcescens [18].

S. marcescens flhDC transcription is negatively and directly regulated by the RssAB two-component system [33] and RcsB [18] and positively and directly regulated by the cyclic-AMP-receptor protein (CRP) [34, 35]. Sulphate was also demonstrated to have a positive effect on flhD expression in S. marcescens through an as yet unknown mechanism [36].

With respect to the role of flhD in the pathogenic capacity of S. marcescens , a flhD mutant was found in a forward genetic screen to be defective in inducing apoptosis in silkworm haemocytes and required a greater than 10-fold higher inoculum for lethality when injected into silkworms in comparison to the wild type [37]. The RssAB-FlhDC signalling pathway was shown to regulate pathogenesis in an animal model [16]. Lin and colleagues reported reduced expression of shlBA and haemolysis in an flhDC mutant, and the opposite phenotype when flhDC was over expressed in S. marcescens strain CH-1 [16]. Cytotoxicity of S. marcescens CH-1 cells to, and invasion into, a human bronchial epithelial cell line, BEAS-2B, was significantly reduced in the flhDC mutant compared to the wild type [16]. These phenotypes could be enhanced by induced expression of flhDC in wild-type CH-1, but not in an isogenic shlBA mutant [16]. Di Venanzio et al. demonstrated that flhD and shlB were defective in haemolysis [18]. However, apart from ShlA, the role of other FlhDC-regulated proteins such as flagellin and phospholipase A protein in cytotoxicity to human epithelial cells has not previously been determined.

It was previously demonstrated that in S. marcescens strain PIC3611, mutation of the crp gene severely reduced expression of flhDC, yielding cells without flagella [34]. A suppressor screen for crp mutant phenotypes resulted in identification of the scrp-31 mutant strain, which has a transposon insertion upstream of the flhDC operon [34]. Strains with the scrp-31 mutation upstream of flhDC were hyper-flagellated and had ~10× higher levels of flhD RNA [34]. In the present study we took advantage of the elevated flhDC expression levels in the scrp-31 strain to investigate the importance of FlhDC-regulated genes in cytotoxicity to human airway and ocular surface epithelial cells in vitro, and used a genetic approach to determine whether PhlA and ShlA contribute to this phenotype.

Methods

Strains and culture media

Escherichia coli and S. marcescens were grown in lysogeny broth (LB) (per litre: 5 g NaCl, 5 g yeast extract, 10 g tryptone) or on LB plates supplemented with 10 g agar l⁻¹. Swimming plates consisted of LB medium supplemented with agar to 0.3 % (w/v). Bacterial cultures were grown on a tissue culture roller drum (TC-7; New Brunswick) at 30 °C, or on agar plates at 30 or 37 °C. Kanamycin was used at 50 µg ml⁻¹ for E. coli and 100 µg ml⁻¹ for S. marcescens ; gentamicin and tetracycline were used at 10 µg ml⁻¹. Saccharomyces cerevisiae, strain InvSc1, was grown in yeast extract-peptone-dextrose medium or on synthetic complete medium without uracil agar for selection of plasmids [38]. All bacterial culture components were from BD Difco, with the exception of antibiotics, which were acquired from Sigma-Aldrich.

For growth curves, single colonies were grown in 5 ml of LB in test tubes overnight at 30 °C (n=4). The cultures were subcultured to an OD₆₀₀ of 0.5 and at various time point aliquots were removed for c.f.u. and optical density evaluation. The experiment was repeated with similar results.

Transmission electron microscopy (TEM)

Cultures were grown overnight as noted above, washed with PBS, applied to Formvar-coated grids, and stained with uranyl acetate (1 %) as previously described [34, 35]. Grids were viewed with a JEM-1210 electron microscope.

Construction of plasmids and strains

All genes were amplified using a high-fidelity polymerase (Phusion; New England Biolabs), and cloned using Saccharomyces cerevisiae-based recombination [39, 40]. Plasmids and oligonucleotide primers are listed in Tables 1 and 2, respectively. Genetic nomenclature followed the guidelines of Berlyn et al. [41].

Table 1.

Strains and plasmids used in this study

Strain or plasmid	Description	Reference or source
Saccharomyces cerevisiae and Escherichia coli
InvSc1	MATa/MATa leu2/leu2 trp1-289/trp1-289 ura3-52/ura3-52 his3-∆1/his3-∆1	Invitrogen
SM10 λpir	thi thr leu tonA lacY Sup.E recA::RP4-2Tc::Mu pir	[72]
S17-1 λpir	thi pro hsdR hsdM + ∆recA RP4-2::TcMu-Km::Tn7 pir	[72]
JM2125 EC100D	[araD139]B/r, eda-10, flhD5301, relA1, rpsL150(strR), deoC1 E. coli stock centre F- mcrA Δ(mrr-hsdRMS-mcrBC) φ80dlacZΔM15 ΔlacX74 recA1 endA1 araD139 Δ(ara, leu)7697 galU galK λ- rpsL (StrR) nupG pir-116(DHFR)	Epicentre
Serratia marcescens
PIC3611	wild type	Presque Isle Cultures
CMS711	scrp-31 mutation in PIC3611, upregulated flhDC	[34]
CMS1324	PIC3611 with shlA::pMQ187	This study
CMS2381	PIC3611 with flhD::pMQ316	This study
CMS2384	CMS711 with flhD::pMQ316	This study
CMS2386	PIC3611 with phlA:: pMQ215	[42]
CMS2388	CMS711 with phlA:: pMQ215	This study
CMS3845	CMS711 with shlA::pMQ187	This study
CMS3901	CMS711 with fliC::pMQ461	[19]
Plasmids
pStvZ3	oriR6K lacZ nptII, oriT, URA3, CEN6/ARSH4	[35]
pMQ118	nptII, rpsL, oriT, oriR6K, URA3, CEN6/ARSH4	[40]
pMQ125	p15a, P BAD -lacZa, ori pRO1600	[40]
pMQ131	oripBBR1, aphA-3, oriT, URA3, CEN6/ARSH4	[40]
pMQ132	oripBBR1, aacC1, oriT, URA3, CEN6/ARSH4	[40]
pMQ164	pMQ118+aacC-1 internal fragment	This study
pMQ187	pMQ118+shlA internal fragment	This study
pMQ192	pMQ118+dtomato	This study
pMQ210	pMQ125+flhDC (arabinose inducible flhDC)	This study
pMQ215	pMQ118+phlA internal fragment	[42]
pMQ236	oriR6K, nptII, rpsL, oriT, URA3, CEN6/ARSH4, I-SceI site	[40]
pMQ240	oripSC101ts, aacC1, oriT, P lac -I-SceI, URA3, CEN6/ARSH4	[40]
pMQ316	pStvZ3+flhD internal region	This study
pMQ461	pMQ192 with fliC internal region	This study
pMQ482	pMQ125+phlAB	This study
pMQ492	pMQ125+shlBA	[25]

Table 2.

Oligonucleotide primers used in this study

Primer no.	Primer sequence*
1022	gacgttgtaaaacgacggccagtgccaagcttgcatGCATCAACATCATCAACACCG
1023	gataacaatttcacacaggaaacagctatgaccatgatGCTGGAGGTGTGGTTGCGCTG
1042	cgacgttgtaaaacgacggccagtgccaagcttgcatGGACCAGTTGCGTGAGCGC
1043	gataacaatttcacacaggaaacagctatgaccatgatGCTTGTAAACCGTTTTGTG
1278	ctcgcatggggagaccccacactaccatcggcgctggTTACTTGTACAGCTCGTCCATGC
1279	tggaattgtgagcggataacaatttcacacaggaggatccaaaaATGGTGAGCAAGGGCGAGGAG
1364	tcagaccgcttctgcgttctgatttaatctgtatcaCAAATCGCGCGTTTAACCTGCTCG
1365	aactctctactgtttctccatacccgtaggaggaaaaaATGGGGAATATGGGTACGTCTG
2119	acgacgttgtaaaacgacgggatctatcatcgtggatccTAGCAAAATGCCGGTATGGAT
2120	ctagagcggtttcccgactggaaagcgggcagtgagcgcCTTAAGCATATTTATGACATC
3302	tgatgacctcctcgcccttgctcaccatttttggatcTCCTGTGCCAGCGCTTTGTCCAG
3303	gttgtgtggaattgtgagcggataacaatttcacacCGTCTGTCTTCCGGTCTGCGTATC
3446	gaattgggtaccgggccccccctcgaggtcgacggtatCAGGCGGTTTTTTTATGGGCGT
3447	ctctctactgtttctccatacccgtaggaggaaaaaATGAGTATGCCTTTAAGTTTTACC

*Sequences are listed 5′ to 3′. Lower case base pairs target recombination and upper case base pairs direct amplification; underlined letters indicate an introduced restriction site.

To mutagenize flhD, fliC, phlA and shlA, internal fragments of these genes from S. marcescens strain PIC3611 were amplified and cloned into suicide plasmid pMQ118, or lacZ or dtomato derivatives, pStvZ3 or pMQ192. Primers 1022 and 1023 were used to amplify an internal fragment from the shlA gene, and the resulting plasmid pMQ187 is designed to integrate at base pair 1395 out of the 4827 bp ORF. pMQ316 was generated by amplifying an internal region of flhD from PIC3611 with primers 2119 and 2120 and for recombination of the amplicon into pStvZ3. The pMQ316 plasmid integrates at base pair 297 out of 360. Primers 1278 and 1279 were used to amplify the dtomato gene and replace the lacZα ORF on pMQ118 generating pMQ192. The fliC insertion plasmid, pMQ461, was made by cloning an internal fragment of fliC amplified with primers 3302 and 3303 into pMQ192. pMQ461 integrates at base pair 811 out of the 1053 bp fliC ORF. The plasmid to mutate phlA, pMQ225, was previously described [42]. These plasmids were introduced into S. marcescens by conjugation as previously described [34, 35]. Kanamycin-resistant isolates were tested for insertion of the plasmid by PCR and phenotypic analysis. As a control for plasmid insertion, an internal fragment of the aacC-1 gene from the mariner transposon from pBT20 was cloned in pMQ118 using primers 1042 and 1043, so that the resulting plasmid, pMQ164, can be integrated into the transposon-associated aacC-1 gene in the scrp-31 mutant chromosome.

For induced expression of genes, fhlDC, phlAB and shlBA operons were amplified from S. marcescens strain PIC3611 or clinical isolate K904, and placed under control of the E. coli P_BAD promoter on plasmid pMQ125. The flhDC operon was amplified using primers 1364 and 1365. The phlAB operon was amplified using primers 3446 and 3447. The shlBA operon expression plasmid, pMQ492, was previously described [25]. Plasmids were verified by PCR, sequencing and phenotypic analysis.

Cell culture and cytotoxicity analysis

A549 human airway carcinoma cells [43] were maintained in Gibco medium number 199 supplemented with 10 % fetal bovine serum (FBS), 5 % sodium bicarbonate, 100 IU ml⁻¹ penicillin, 100 µg ml⁻¹ streptomycin, 100 µg ml⁻¹ gentamycin and 2.5 µg ml⁻¹ amphotericin B. When cells were plated for these experiments, outgrowth media for A549 cells consisted of Gibco medium 199 with 10 % FBS and 5 % sodium bicarbonate, but without antibiotics. Human corneal limbal epithelial (HCLE) cells [44] were grown in keratinocyte serum-free medium (KSFM) with l-glutamine (Invitrogen), supplemented with 25 μg ml⁻¹ bovine pituitary extract, 0.2 ng ml⁻¹ epidermal growth factor and 0.3 mM CaCl₂, without antibiotics.

Cytotoxicity was measured using confluent cell layers in 24-well dishes exposed to filtered bacterial supernatants or bacteria at an m.o.i. of 200 or 20 as previously described [45]. For bacterial cell cytotoxicity assays, bacteria were grown for 18 h in LB with aeration, adjusted to an OD₆₀₀ of 1, washed in PBS, and suspended in tissue culture medium to the indicated m.o.i. Confluent epithelial cell layers were challenged with bacteria for 2 h at 37 °C in 5 % CO₂. Bacteria were removed by aspiration and cell layers were gently washed three times with warm PBS. Growth medium OG or KSFM with antibiotics were then added to the cell layers with a vital stain, resazurin (Presto Blue), as suggested by the manufacturer (Invitrogen) or Calcein AM (0.5 µM). Cell layers were stained for 60 min (resazurin) or 15 min (Calcein AM) before processing. For bacterial secreted secretome cytotoxicity assays, bacteria were grown as above, then adjusted to an OD₆₀₀ of 2.0 with fresh LB and removed with centrifugation and filtration (Millex GV 0.22 µm; Millipore). Monolayers with 300 µl of OG or KSFM were challenged with 200 µl of filtered supernatants and incubated for 4 h at 37 °C in 5 % CO₂. Viability was compared using mock-treated samples, which were treated with an equivalent amount of PBS (for bacterial challenge experiments) or LB for secretome experiments. Mock-treated samples were set at 100 % and cells treated with 0.25 % Triton X-100 for 10 min were set at 0 % viability. Relative viability was determined with the following equation: 100× [1 – (maximum value – experimental value)/(maximum value – minimum value)], with mock-treated cells being the maximum value and Triton X-100-treated cells the minimum value.

For fluorescence images, cells on glass-bottomed multiwell plates (MatTek) were imaged with a 10× objective using an Olympus IX-81 inverted microscope equipped with an FV-1000 laser scanning confocal system (Olympus) with FluoView FV10-ASW 3.1 imaging software. For brightfield images, samples in multiwell dishes (Costar 3526) were viewed with a Nikon TE2000-E microscope equipped with a Photometrics CoolSNAP HQ-camera and a 10×0.30 NA objective. NIS-Elements 3.2 software was used to obtain digital images.

Galleria mellonella infection model

Using the basic protocol of previous studies, virulence of various bacterial strains were tested in a G. mellonella model of pathogenesis [25, 46]. Bacteria were grown for 18 h in LB medium at 30 °C, pelleted by centrifugation, washed and resuspended into PBS solution, normalized to an OD₆₀₀ of 0.5 and serial diluted using 10-fold intervals. Each dilution was injected into 10–15 G. mellonella larvae (GrubCo), and plated on LB agar plates to determine the actual c.f.u. injected. A negative control group of 10–15 larvae were injected with PBS without bacteria; at least 90 % of these survived in each experiment. Each group of 10 larvae were placed in an empty Petri dish and incubated at 30 °C. After 18–20 h the larvae were assessed for viability; those that were motile when prodded gently with a blunt pencil tip were counted as alive, and those that were deep red or black and non-motile were counted as dead. Larvae were placed at −80 °C for 1 day and then autoclaved to kill and sterilize living larvae and bacteria. These experiments were repeated four times with similar results, using different batches of larvae.

Statistical analysis

All experiments were performed at least twice with independent biological replicates. Significance was set at P≤0.05. Prism software (Graphpad) was used to graph and analyse data with two-tailed Student’s t-tests and one-way ANOVA with Tukey’s post-hoc test. Fisher's exact test was used with an online contingency table calculator at graphpad.com.

Results

The FlhDC transcriptional regulator mediates cytotoxicity

The previously identified scrp-31 mutation in S. marcescens strain PIC3611 increases expression of the flhDC operon by approximately 10-fold compared to expression in the wild-type strain due to the insertion of a transposon with a strong promoter upstream of the flhDC operon [34]. This manifests in pleiotropic phenotypes including increased production of flagella relative to the wild type (compare Fig. 1a, b) and a slight reduction in growth (Fig. 1c). In stationary phase cultures (t=24 h), c.f.u. values were only slightly reduced, with 1.6×10¹⁰±3.2×10⁹ in the wild-type culture and 9.0×10⁹±5.5×10⁹ in the scrp-31 culture (P=0.0091).

Fig. 1.

The scrp-31 mutation leads to increased FlhDC-mediated phenotypes including increased flagellum production and a slight reduction in growth rate. (a, b) Representative TEM micrographs of bacteria grown to stationary phase. Bar, 500 nm. (a) WT. (b) scrp-31. (c) Growth curve in LB media (n=8). Mean and sd are shown.

In preliminary studies, this scrp-31 mutant strain was found to be highly cytotoxic to human cell lines in vitro. This was a striking finding because in a previous study the parental strain PIC3611 induced lower cytotoxicity than 26 other tested strains to an ocular surface cell line [47]. The present study was focused on identifying the mechanism by which the scrp-31 strain was more cytotoxic than the isogenic wild-type strain with the larger goal of understanding the role of FlhDC in S. marcescens host–pathogen interactions.

HCLE cell lines were challenged with bacteria at an m.o.i. of 200 to test the cytotoxicity of bacteria. After 2 h of co-incubation, the bacteria were removed by washing, the culture medium was replaced including antibiotics to kill any remaining bacteria, and culture viability was measured using a quantitative viability dye (resazurin). Whereas the cells of the wild-type bacterial strain PIC3611 were not very cytotoxic to the HCLE cells, the scrp-31 mutant bacteria were highly cytotoxic (Fig. 2a). Similar results were observed with an m.o.i. of 20 (data not shown).

Fig. 2.

Overexpression of flhDC in S. marcescens confers elevated cytotoxicity to human corneal epithelial cells in vitro. Plus and minus symbols indicate expected FlhD activity. (a) Analysis of HCLE monolayers exposed to S. marcescens (m.o.i.=200) for 2 h, followed by removal of the bacteria, and viability staining with resazurin. The average and sd are shown; n=4 from four independent experiments each with three wells. *Significantly different from the WT group by ANOVA with Tukey’s post-hoc test. (b) Calcein AM viability staining of HCLE cells exposed to S. marcescens as in (a), washed to remove bacteria, then stained and imaged with identical exposures. Positive staining indicates viability; a representative image at 100× magnification is shown. Mock indicates no bacterial challenge. (c) Induced expression of flhDC from a plasmid (pflhDC) in the WT (PIC3611) increased cytotoxicity to HCLE cells in vitro as tested in (a). Vector, pMQ125; pflhDC, pMQ210. The mean and standard deviation are shown (n=4). *Significantly different from the WT+vector group by Student’s t-test with Yates correction.

As a second method, calcein AM vital stain was used to verify cytotoxicity of the scrp-31 mutant bacteria to the HCLE cell line (Fig. 2b). Together, these assays support that elevated expression of flhDC confers a hyper-cytotoxic phenotype to S. marcescens .

To confirm that this effect was due to flhDC expression rather than some unknown mutation elsewhere in the chromosome, the flhD gene in the scrp-31 mutant strain was mutated. The flhD mutation eliminated the hyper-cytotoxicity phenotype of the scrp31 mutant, confirming the role of flhDC in the hyper-cytotoxic phenotype (Fig. 2a, b). Mutation of flhD in the parental strain, PIC3611, also reduced cytotoxicity (Fig. 2a); viability levels were slightly higher than the mock treatment (no bacteria challenge), which was set at 100 % and used to establish relative viability (see Methods). This unexpected increase was unlikely to be due to the presence of residual bacteria, because after the 2 h period in which the HCLE cells were challenged with bacteria, they were washed three times and then incubated in growth medium with antibiotics effective against the bacteria. Additionally, c.f.u. analysis of the medium did not reveal viable bacteria (limit of detection=100 bacteria, data not shown).

To further support that elevated expression of flhDC led to cytotoxic effects, we cloned the flhDC operon from S. marcescens and placed it under control of the P_BAD promoter on a multicopy plasmid. Induced expression of flhDC in the wild-type S. marcescens strain PIC3611 conferred an increase in cytotoxicity to the HCLE cells (Fig. 2c). These data support the idea that increased FlhDC can regulate cytotoxicity to human epithelial cells.

ShlA is necessary for FlhDC-mediated cytotoxicity of bacterial cells to an ocular and airway cell line

A reverse genetics approach was used to determine the mechanism of FlhD-mediated cytoxicity. Because FlhDC positively regulates transcription of phospholipase (phlA) and the haemolysin operon (shlBA), both of which have been implicated in lysis of red blood cells, in addition to its better known role in flagella biosynthesis, we tested the hypothesis that phlA, shlBA and flagellin genes are responsible for the hyper-cytotoxic phenotype of the scrp-31 mutant strain. Double mutants of scrp-31 fliC, scrp-31 phlA and scrp-31 shlA were generated and tested for cytotoxicity. The fliC gene codes for the flagellin subunit.

As noted above, HCLE cells were challenged with the scrp-31 mutant with a control insertion in the aacC-1 gene of the transposon upstream of flhDC that creates the scrp-31 phenotypes at an m.o.i. of 200. This plasmid did not alter scrp-31 phenotypes as expected (data not shown), but served as a control to ensure that the plasmids inserted in fliC, phlA and shlA did not confer unexpected phenotypes. Similar to the results shown in Fig. 2a, the scrp-31 mutant with the vector inserted in the transposon was highly toxic to HCLE cells (Fig. 3a, b). Cell layers lost their cobblestone appearance, became flattened and developed large vacuoles. Like scrp-31, the fliC and phlA double mutants were highly toxic to the HCLE cells (Fig. 3a, b). Strikingly, the scrp-31 shlA double mutant was non-cytotoxic, similar to the scrp-31 flhD double mutant (Fig. 3a, b).

Fig. 3.

FlhDC-mediated bacterial cytotoxicity to a human ocular and airway cell line required ShlA. HCLE and A549 monolayers exposed to bacteria (m.o.i.=200) for 2 h. Bacteria were removed by gentle washing and mammalian cells were assessed by (a) phase contrast microscopy, (b) viability staining of HCLE cells with resazurin, and (c) viability staining of A549 cells with resazurin. Mock indicates no bacterial challenge. (a) Arrows indicate example vacuoles. For (a), a representative experiment is shown, for (b), an average of six wells from two separate experiments is shown, and for (c), an average of nine independent experiments is shown. For (c), Vector=pMQ125, pflhDC=pMQ210, pphlAB=pMQ482, and pshlBA=pMQ492. Error bars indicate one standard deviation. *Significant reduction compared to the WT (PIC3611); ANOVA with Tukey’s post-hoc test.

To test whether this phenotype extends beyond ocular cells, the human airway carcinoma cell line A549 was used. A similar pattern emerged with the scrp-31 mutant being highly toxic, and mutation of flhD and shlA suppressed cytotoxicity (Fig. 3a). The impact on A549 cells appeared to be more severe, with cells releasing from the surface of the plate and becoming transparent (Fig. 3a). Together, these data suggest that the mechanism of FlhDC-mediated bacterial toxicity to ocular cells is independent of the flagellum and phospholipase A and dependent upon the ShlA haemolysin.

To validate the importance of expression of FlhDC and ShlA in bacterial cytotoxicity, the flhDC, phlAB and shlBA operons were cloned under transcriptional control of the P_BAD promoter on a multicopy plasmid. These plasmids were used to transform strain PIC3611 (WT). The transformed bacteria were induced with arabinose, and the resulting bacteria were used to challenge A549 cells. Induction of flhDC and shlBA, but not the vector negative control or phlAB, conferred high levels of cytotoxicity to A549 cells (Fig. 3c). Together, these data support that FlhDC-mediated bacterial cytotoxicity to human epithelial cells is dependent on ShlA.

ShlA and PhlA are necessary for FlhDC-mediated cytotoxicity of bacterial secreted factors to an airway cell line

We tested whether the secreted supernatant, or secretome, was cytotoxic to human epithelial cell lines in the absence of bacteria. Unlike bacterial cells, the secretome of scrp-31 mutants produced a non-significant ~20 % loss in viability of HCLE cells in vitro using the resazurin tests that was indistinguishable from the effect of WT secretomes (P=0.56) (Fig. 4a, b). However, the scrp-31 secretome was highly toxic to the A549 cell line (Fig. 4a, b). Whereas mutation of fliC did not significantly affect the cytotoxicity of scrp-31 mutant secretomes (Fig. 4a, c), mutation of flhD, phlA and shlA eliminated the cytotoxic phenotype conferred by the scrp-31 mutation. Together, these results indicate that both PhlA and ShlA are important for FlhDC-mediated cytotoxicity secreted by S. marcescens .

Fig. 4.

FlhDC-mediated bacterial cytotoxicity to a human ocular and airway cell line required PhlA and ShlA. HCLE and A549 monolayers were exposed to bacterial secretomes from stationary phase bacterial cultures and normalized to an OD₆₀₀ of 2.0, for 4 h. Supernatants were removed by gentle washing and cell layers were assessed by (a) phase contrast microscopy (representative images are shown), and (b, c) viability staining with resazurin of HCLE (b) and A549 (c) cells (n≥5 per group from six independent experiments). Mean and sd are shown. *Significant reduction compared to the WT (PIC3611), ANOVA with Tukey’s post-hoc test. Error bars indicate one standard deviation. Mock indicates no bacterial challenge.

FlhDC and ShlA mediate S. marcescens virulence to G. mellonella larvae

The importance of FlhDC and ShlA was tested in a multicellular organism using a G. mellonella invertebrate pathogenesis model [46, 48]. S. marcescens wild-type strain PIC3611 is highly virulent to G. mellonella larvae, with an LD₅₀ of less than 10 bacteria per larvae by 24 h (data not shown). In a representative experiment, when 100 c.f.u. bacteria were injected per larvae (Fig. 5a), by 22 h, 100 % of larvae with the scrp-31 mutant died, whereas larvae infected with the scrp-31 flhD and scrp-31 shlA double mutants were less able to kill the larvae with 36 and 17 % killing respectively (n=12 larvae per genotype). PBS-injected control larvae survived. The difference between scrp-31 and the double mutants was significant (Fisher’s exact test, P<0.05). Mutation of phlA did not influence lethality of the scrp-31 strain (Fig. 5a).

Fig. 5.

FlhD and ShlA are required for virulence in a G. mellonella infection model. (a) Survival of larvae infected with 100 c.f.u. of scrp-31 and derivative strains at 22 h post-injection (n=12). (b) Survival rate of G. mellonella larvae 19 h post-injection with different doses of bacteria (n=15 larvae per genotype and dose) from the wild-type background and derived mutants.

To validate the findings, the wild-type strain was used in order to evaluate the role of FlhDC and ShlA without the effect of highly over-expressed flhDC (Fig. 5a). When the wild type and isogenic flhD and shlA mutants were injected into larvae there was a dose-dependent killing effect for all strains, but the flhD and shlA strains were clearly less virulent (Fig. 5b). Strains with mutations in phlA were equally virulent to the parental wild-type strain when injected into larvae. Larvae injected with PBS vehicle showed no loss of viability (data not shown).

Discussion

During the course of experimentation, we noted that scrp-31 mutants of S. marcescens strain PIC3611 were highly toxic to epithelial cells, leading to this study to determine the responsible mechanism. Experiments here validated that elevated flhDC expression makes the bacterial cells and bacterial secretomes more cytotoxic to human epithelial cell lines. This conclusion was based on the correlation between elevated toxicity and flhDC expression in the scrp-31 mutant strain and in the PIC3611 strain with multicopy expression of flhDC; moreover, mutation of flhD in both the wild type and the scrp-31 mutant eliminated the hyper-cytotoxicity phenotype. The lower cytotoxicity derived from plasmid-based expression of flhDC relative to the scrp-31 mutant may stem from a weaker promoter, as the strong P_tac was used to drive flhDC expression in scrp-31, and the relatively weak P_BAD being used with the plasmid. Our findings agree with an elegant previous study, where an flhDC mutant in strain CH-1 was observed to be less cytotoxic to a human bronchial epithelial cell line [16], but this defect was not functionally complemented by expression of downstream virulence factors. This study supplies the next step by testing the individual virulence factors downstream of FlhDC with respect to their role in cytotoxicity to human cells.

Whereas the flagellin mutant was equally cytotoxic to the wild type, this was not a forgone conclusion, as flagellin has been shown to induce inflammation and apoptosis in gastrointestinal cells through caspase activation pathways [49, 50] and pyroptosis in neutrophils and macrophages in an NLRC4 inflammasome-dependent manner [51, 52]. In this study the lack of a role for flagellin may result from the relatively short duration of the cytotoxicity assays in this study, differences in cell types, or because the flagellin may not harm the tested cell types. Indeed, a protective role for flagellin for corneal cells has been reported [53–55].

It was not previously tested whether bacteria lacking phlA were toxic to human airway or ocular cells, although supernatants from stationary phase cultures of phlA mutants were shown to be less cytotoxic to HeLa cells [24]. Here we found that PhlA was not required for cytotoxicity of bacterial cells to epithelial cells; however, PhlA does contribute to the cytotoxicity of bacterial secreted products to human airway cells, expanding the range of known PhlA-sensitive cells beyond erythrocytes and HeLa cells. It is not clear why PhlA is not required in bacterial cytotoxicity. It may be that the 2 h exposure of cells used for the bacterial cytotoxicity model is not sufficient to generate cytotoxic levels of PhlA. In contrast, PhlA is capable of catalysing the production of the cytolytic biosurfactant lysophospholipid from mammalian phospholipids [24], but phlA mutants were not altered in haemolytic activity in one study [18]. Consistent with a role for PhlA in pathogenesis, an in vivo study demonstrated that phlA mutants had a mild reduction in fitness relative to wild-type bacteria in a murine urinary tract infection model [56]. The S. marcescens phlA gene has been demonstrated to be positively regulated by FlhDC in E. coli and in S. marcescens [15, 16].

Of note, the airway epithelial cells were far more susceptible to secreted factors than the ocular cells, and this was dependent upon both PhlA and ShlA. While it would be premature to conclude that all airway cells are susceptible and all ocular surface cells are resistant to secreted bacterial proteins such as PhlA based on one cell line per tissue type used in this study, differences in cell type susceptibility to PhlA have been noted by other groups, suggesting that different cell types have variable susceptibility. For example, Shimuta et al. observed that human, but not horse or sheep, erythrocytes are damaged by S. marcescens PhlA [24]. Furthermore, ocular surface cells are continually exposed to human secretory phospholipase A(2), which is a major antimicrobial component of human tears and is present at around 50 µg ml⁻¹ in normal tears, suggesting that ocular surface cells are tolerant of phospholipase A enzymes [57, 58].

ShlA has been shown to be cytotoxic to a variety of human cell lines, including epithelial cells, endothelial cells, fibroblasts and immune cells [16, 22, 23]. As we observed here with bacteria expressing shlBA from a plasmid or with elevated flhDC expression levels, a previous study reported that the addition of purified ShlA to epithelial cells induced production of vacuoles in a manner with similarities to the action of the VacA from Helicobacter pylori and aerolysin from Aeromonas hydrophilia [22].

Interestingly, we found that ShlA was important not just for bacterial cytotoxicity but also for secretome cytotoxicity. Whereas it is sometimes noted as a largely surface-associated protein, evidence suggests that ShlA can be found in culture supernatants [59, 60], supporting that ShlA acts as both a secreted and surface-associated cytolysin.

In this study, ShlA was the key determinant of FlhDC-mediated cytotoxicity, as the flhD mutant cytotoxicity and virulence were virtually eliminated. This suggests that other regulators of ShlA such as RcsB [31] do not facilitate shlBA expression during the in vitro and in vivo conditions used in this study. RcsB is a negative regulator of shlBA and may actually shut down expression of shlBA in the G. mellonella larvae because it can be activated by antimicrobial peptides [61] such as those found in G. mellonella haemolymph [62].

In Serratia , beyond regulating cytotoxicity and motility, FlhDC controls biofilm formation on abiotic surfaces in an ShlBA-independent manner [16]. FlhDC-mediated attachment to abiotic surfaces is probably mediated by regulation of type I pili [19]. Nevertheless, ShlA is important for invasion into mammalian cells [16, 23] and may facilitate attachment to cells. The bacterial capsule or enterobacterial common antigen (ECA), a surface glycolipid, may also play a part in FlhDC-mediated attachment, as one study demonstrated co-regulation of flhD and the ECA biosynthesis genes, such as wecD [12]. This gene cluster is also predicted to be necessary for colanic acid capsule production, which has been shown to have a role in early biofilm formation by E. coli [63].

The major limitations of this study include that experiments were performed in vitro, and that only one strain background was used. These studies will have to be validated using animal models of infection and genetically modified clinical isolates of S. marcescens .

In summary, this study determined that both PhlA and ShlA contribute to the elevated cytotoxicity of the scrp-31 mutant and demonstrate differences between airway and ocular surface cell lines in susceptibility to S. marcescens secretomes and PhlA in particular. Importantly, S. marcescens is an important pathogen for hospital-acquired infections [64] and contact lens-associated keratitis, for which it is second only to Pseudomonas aeruginosa in frequency of isolation [65–68]. Additionally, S. marcescens is a common contaminant of contact lens cases and lenses [69], such that secreted products from the bacteria could be exposed to the eye via the lens with or without bacterial attachment to the lens, leading to conditions such as sterile infiltrates and contact lens-associated red eye (CLARE). Moreover, because of the high level of susceptibility of human airway cells to S. marcescens and the increased use of ventilators during the COVID-19 pandemic that may lead to elevated levels of ventilator-associated bacterial pneumonia [70, 71], understanding the mechanisms of bacterial virulence is more relevant than ever.