Biologically active pigment and ShlA cytolysin of Serratia marcescens induce autophagy in a human ocular surface cell line

Bacterial stocks (Table 1) were stored at − 80 °C and single colonies were obtained on lysogeny broth (LB) agar. Colonies were grown in LB at 30 °C for ~ 18 h with aeration on tissue culture rollers. Where noted, bacteria were grown with L-arabinose at 1 mM for controlled expression of genes. Secretomes were prepared by normalizing overnight cultures to OD₆₀₀ = 2.0, removal of bacteria by centrifugation at 14,000 rpm and filtration through a 0.22 μm filter (Millex PVDF). Normalized secretomes were added to HCLE cells at a ratio of 500 μl per 1 ml of tissue culture medium (KSFM) and incubated at 37 °C + 5% CO₂ for 3 h. In some cases secretomes were further diluted 2-fold (OD₆₀₀ = 1.0) due to excessive cytotoxicity as noted in the text. The autophagy inhibitor 3-methyladenine (3MA) was added to culture media at 5 mM, one hour prior to challenge with secretomes as previously described [18, 31].

To analyze autophagy, LC3-GFP HCLE cells [18] were imaged with 60X magnification on an Olympus IX-81 inverted confocal microscope with Fluoview imaging software. The LC3-GFP cells were generated by lentiviral transduction of the human corneal limbal epithelial cell line from Ilene Gipson [32]. Image J (NIH) was used to quantify images without any image adjustment. Autophagy levels were quantified following recommendations of Klionsky et al. [33] in which the standard deviation of fluorescent pixel intensity of a cell is divided by its mean pixel intensity of the cell. Two to three fields per treatment condition were imaged. The experiment was repeated on at least two different days and at least 50 cells were analyzed per group. The data was averaged from all fields taken per experiment and graphed using GraphPad prism. One way ANOVA with Tukey’s post hoc analysis was used to determine statistical significance at P < 0.05 unless otherwise stated.

Additional mutants (Table 1) were generated by transposon mutagenesis using the pSC189 mariner transposon delivery system as previously described [34, 35]. Transposons were mapped using the method of Chiang, et al. [34] To identify nuclease defective mutants, libraries of mutants were transferred from 96-well plates to DNase detection agar (BD Difco) and plates were screened for loss of nuclease zones after 16–20 h of incubation at 30 °C around individual colonies. The transposon was mapped to 89 bp upstream of the nucA open reading frame and results in an almost complete loss of secreted nuclease activity (data not shown). Mutations in the prodigiosin biosynthetic operon were obtained by visually screening mutant libraries of strain K904 and K904 ∆shlB for loss of pigment. Transposon insertions were mapped to base pair 2451 in the K904 strain background and to base pair 1075 of pigE in the ∆shlB strain background.

Wild-type strain PIC3611 and an isogenic ∆pigA strain were grown overnight in LB broth with aeration. The ∆pigA mutant does not make prodigiosin and served as a negative control. Bacteria were adjusted to OD₆₀₀ = 4, aliquots (5 ml) were pelleted by centrifugation (7000 RPM for 10 min), and supernatants were removed. To extract prodigiosin, bacterial cells were suspended in 100 μl of acidified ethanol (2 ml of 2 M HCl in 98 ml of 95% ethanol) and incubated for one hour with periodic vortexing. Samples were further purified with hydroxyapatite resin. Columns were packed with hydroxyapatite resin (BioRad #16260), equilibrated with acidified ethanol, and samples were run through the columns with acidified ethanol. The mock purification sample from the ∆pigA culture was collected at the same time as the prodigiosin fraction from the wild type was collected. Samples were air-dried and prodigiosin concentration was determined using a standard curve of absorbance at 534 nm using commercial prodigiosin as a standard (Sigma). The same volume of prodigiosin was added from the wild type and ∆pigA-derived samples.

Our previous study showed that a subset of ocular bacterial pathogens induced autophagy in a corneal cell line, and that among the strongest induction was observed with S. marcescens [18]. In this study, we set out to determine which components of S. marcescens using wild-type and genetically manipulated strains induced formation of LC3-GFP puncta. We used two S. marcescens strains: strain PIC3611 from Presque Isle Culture collection, a laboratory strain that is likely from an environmental source (biotype TCT), and K904, a contact lens associated keratitis isolate (biotype A6a). Two strains were used to determine whether phenotypes were associated with one particular strain or a more general phenomenon.

Figure 1 demonstrates activation of autophagy following exposure of HCLE LC3-GFP cells to normalized filtered supernatants (secretomes) from strain PIC3611 Fig. 1a, b). The formation of LC3-GFP puncta indicate cells with activated autophagy and can be used in quantification of autophagy [33]. Autophagy stimulation by PIC3611 secretomes could be prevented using autophagy inhibitor 3-methyladenine (3MA), supporting that the observed LC3-GFP phenotype is autophagy dependent (Fig. 1). The same trend was previously shown for strain K904 and with the use of autophagy inhibitor baflomycin [18]. As an additional control to show that LC3-GFP puncta are not an artifact of fluorescent bacterial components, an HCLE-GFP cell line with no LC3 fusion was used [18]. Following PIC3611 supernatant treatment, no fluorescent focus formation was observed from the GFP control cell line (Fig. 1c).

The component(s) of the secretome responsible for autophagy induction is unknown. Our previous study demonstrated that secretomes could be heated at 95 °C and still induce autophagy [18]. Well-defined S. marcescens strains with deletion mutations in the eepR and gumB genes were tested for loss of autophagy induction because these genes have a conserved and broad impact on S. marcescens biology [20, 24, 25, 28, 36]. The EepR and GumB genes regulate expression of multiple secreted factors including heat-stable secondary metabolites such as the biologically active pigment prodigiosin and the hemolytic and antimicrobial biosurfactant serratamolide. Both the eepR and gumB mutants were defective in activation of autophagy (Fig. 2a).

Additionally, secretomes from strains with mutations in a variety of surface or secreted proteins regulated by EepR and/or GumB were individually tested in the K904 strain background (Figs. 2b and 3). These include the type 1 pilus gene, fimC, the flagellin gene, fliC, the secreted nuclease gene, nucA, the prodigiosin biosynthetic gene, pigD, the cytolysin, shlA, the outer membrane cytolysin transporter gene shlB, the S-layer protein gene, slaA, and the serratamolide biosynthetic gene, swrW.

Mutant analysis suggested that the eepR and gumB mutants share a defect in autophagy induction with a subset of the mutants (Figs. 2b and 3), most notably pigD, which codes for a gene involved in prodigiosin pigment biosynthesis [37]. Both eepR and gumB are severely defective in prodigiosin production due to reduced transcription of the prodigiosin biosynthetic genes, implicating prodigiosin as a stimulatory factor [25, 28]. Secretomes from mutants unable to make type 1 pili, flagella, and cytolysin / cytolysin transporter (shlA and shlB, respectively) were also defective, but to a lesser extent then the pigment mutants.

Given that the prodigiosin defective mutants, eepR, gumB, and pigD were defective in inducing autophagy whether they were from the PIC3611 (Fig. 2a) or K904 background (Figs. 2a, b and 3), we analyzed whether prodigiosin played a role in inducing autophagy in the HCLE LC3-GFP cell line in greater depth. First, because the eepR mutation confers pleiotropic effects, we compared the wild-type strain PIC3611 with an isogenic mutant unable to make PigA, which is required for prodigiosin biosynthesis. Similar to the eepR mutant, when only pigment biosynthesis was ablated through deletion of the pigA gene, strain PIC3611 was unable to induce autophagy (Fig. 4a).

We have previously described the use of plasmid pMQ262, which replaces the normal pigment biosynthetic promoter with an arabinose inducible promoter [26, 38], such that prodigiosin pigment biosynthesis is dependent upon arabinose in the growth medium (Fig. 4b). When secretomes from strain PIC3611 with pMQ262 were used to challenge HCLE LC3-GFP cells, the ability to activate autophagy correlated with arabinose induction of pigment production (Fig. 4c). Arabinose, itself, did not induce autophagy (data not shown). These data suggest that prodigiosin is necessary for S. marcescens secretomes to fully activate autophagy in corneal cells.

To test whether prodigiosin was sufficient for activation of autophagy in HCLE LC3-GFP cell line, we purified prodigiosin from strain PIC3611 and mock purified it from the isogenic ∆pigA mutant, and tested these for activation of LC3-GFP puncta formation (Fig. 5a-b). Prodigiosin from PIC3611 (0.9 μM) was able to activate autophagy. The negative control mock purified prodigiosin from the ∆pigA mutant was unable to activate autophagy when added at the same volume. Similarly, commercially available prodigiosin could activate autophagy in a dose dependent manner (Fig. 5a, c).

Data from genetic analysis above suggested that the pore forming cytolysin, ShlA, contributes to autophagy induction (Figs. 2b and 3). The ShlB protein is necessary for secretion of ShlA, such that shlB mutants do not secrete ShlA [39]. Consistently, the tested shlB mutant behaved similarly to the shlA mutant (Fig. 2b, and Fig. 3). We generated a double mutant that is unable to make prodigiosin or secrete ShlA (∆shlB pigD), and this was indistinguishable from the pigD mutant, but had a trend to lower levels of autophagy induction compared to the ∆shlB levels.

Our previous study detected moderate activation of autophagy by a clinical keratitis isolate of E. coli in HCLE-GFP cells [18] and that E. coli with a shlBA plasmid (pMQ492) is able to secrete functional ShlA cytolysin [20]. Here, we observed that ectopic expression of the S. marcescens shlBA operon increased the ability of E. coli secretomes to induce autophagy in HCLE LC3-GFP cells (Fig. 6). Together, these results indicate a role for the ShlA cytolysin in activation of autophagy.