ShlA toxin of Serratia induces P2Y2- and α5β1-dependent autophagy and bacterial clearance from host cells

Serratia marcescens is an opportunistic human pathogen involved in antibiotic-resistant hospital acquired infections. Upon contact with the host epithelial cell and prior to internalization, Serratia induces an early autophagic response that is entirely dependent on the ShlA toxin. Once Serratia invades the eukaryotic cell and multiples inside an intracellular vacuole, ShlA expression also promotes an exocytic event that allows bacterial egress from the host cell without compromising its integrity. Several toxins, including ShlA, were shown to induce ATP efflux from eukaryotic cells. Here, we demonstrate that ShlA triggered a nonlytic release of ATP from Chinese hamster ovary (CHO) cells. Enzymatic removal of accumulated extracellular ATP (eATP) or pharmacological blockage of the eATP-P2Y2 purinergic receptor inhibited the ShlA-promoted autophagic response in CHO cells. Despite the intrinsic ecto-ATPase activity of CHO cells, the effective concentration and kinetic profile of eATP was consistent with the established affinity of the P2Y2 receptor and the known kinetics of autophagy induction. Moreover, eATP removal or P2Y2 receptor inhibition also suppressed the ShlA-induced exocytic expulsion of the bacteria from the host cell. Blocking α5β1 integrin highly inhibited ShlA-dependent autophagy, a result consistent with α5β1 transactivation by the P2Y2 receptor. In sum, eATP operates as the key signaling molecule that allows the eukaryotic cell to detect the challenge imposed by the contact with the ShlA toxin. Stimulation of P2Y2-dependent pathways evokes the activation of a defensive response to counteract cell damage and promotes the nonlytic clearance of the pathogen from the infected cell.

Serratia marcescens is an opportunistic human pathogen involved in antibiotic-resistant hospital acquired infections.Upon contact with the host epithelial cell and prior to internalization, Serratia induces an early autophagic response that is entirely dependent on the ShlA toxin.Once Serratia invades the eukaryotic cell and multiples inside an intracellular vacuole, ShlA expression also promotes an exocytic event that allows bacterial egress from the host cell without compromising its integrity.Several toxins, including ShlA, were shown to induce ATP efflux from eukaryotic cells.Here, we demonstrate that ShlA triggered a nonlytic release of ATP from Chinese hamster ovary (CHO) cells.Enzymatic removal of accumulated extracellular ATP (eATP) or pharmacological blockage of the eATP-P2Y2 purinergic receptor inhibited the ShlA-promoted autophagic response in CHO cells.Despite the intrinsic ecto-ATPase activity of CHO cells, the effective concentration and kinetic profile of eATP was consistent with the established affinity of the P2Y2 receptor and the known kinetics of autophagy induction.Moreover, eATP removal or P2Y2 receptor inhibition also suppressed the ShlA-induced exocytic expulsion of the bacteria from the host cell.Blocking α5β1 integrin highly inhibited ShlA-dependent autophagy, a result consistent with α5β1 transactivation by the P2Y2 receptor.In sum, eATP operates as the key signaling molecule that allows the eukaryotic cell to detect the challenge imposed by the contact with the ShlA toxin.Stimulation of P2Y2-dependent pathways evokes the activation of a defensive response to counteract cell damage and promotes the nonlytic clearance of the pathogen from the infected cell.
Serratia marcescens is an opportunistic pathogen that can lead to life-threatening disease such as meningitis, endocarditis, pneumonia, and bacteremia that may lead to sepsis (1-3).The increasing incidence of S. marcescens in clinical settings is due to the expression of various virulence factors, the acquisition of multiple antibiotic resistances, and the ability of the pathogen to resist disinfection procedures (4)(5)(6)(7).The World Health Organization classified S. marcescens among the pathogens that are a research priority to design alternative antimicrobial strategies (8).Therefore, the understanding of the molecular mechanism of action of Serratia virulence effectors will lead to the rational design of novel antibacterial therapies.
S. marcescens expresses and releases the ShlA toxin to the surface of the bacteria, where it can get in contact with host eukaryotic cells (9,10).The shlBA operon encodes the Type Vb two-partner secretion system composed by the ShlB translocator and by ShlA (11,12).As a type V secretion system-delivered effector, ShlA does not show homology to other class of cytolysins such as the repeats in toxin or the cholesterol-depending cytolysins pore-forming toxins (13).ShlA exerts a cytotoxic action on erythrocytes, fibroblasts, and epithelial cells (14,15).In animal models of infection, strains lacking ShlA expression are strongly attenuated in their pathogenic phenotypes (16)(17)(18).
Autophagy (AP) is a key cellular quality control process in eukaryotes being engaged in normal physiology and to counter diverse forms of cellular stress.Although certain microbes are able to hijack the autophagic process to their own benefit, the autophagic response to microbial invaders includes the removal of the pathogen and its virulence effectors (19).We have previously shown that ShlA is able to trigger a reversible autophagic response before Serratia internalization in epithelial cells (20).
After internalization, S. marcescens resides and proliferates in autophagic-like vacuoles and is able to avoid lysosomal elimination (21).Later on, intravacuolar Serratia egresses from the host cell by provoking an ShlA-dependent nonlytic, exocytic mechanism (22), which eliminates the multiplied pathogen from the infected cell but allows bacterial extracellular dissemination.
ShlA was previously shown to induce ATP depletion from epithelial cells and fibroblasts (14), but the mechanisms mediating ATP efflux and the resulting accumulation of extracellular ATP (eATP) on the infection process were not studied.Intracellular ATP (iATP) can be released from different cell types by calcium-regulated exocytosis, membrane transporters, and channels, as well as by cell lysis (23)(24)(25).On the other hand, an important conduit-mediating ATP release is the pannexon, homoheptamer of pannexin1 (PNX1), a protein expressed in the cell membranes of many cell types (26).Depending on the cell type, stimulus, and metabolic status, different ATP conduits can be activated (27).
Most cellular responses to eATP and other nucleotides are mediated by purinergic P2 receptors, classified as P2X and P2Y (28).In vivo, all P2X receptors are exclusively activated by eATP and mediate the transport of Na + , K + , and Ca 2+ across the plasma membrane.eATP can also bind P2Y receptors that activate G proteins, promoting changes in the concentrations of cytosolic calcium and/or cAMP and downstream signaling routes (28).The strength and duration of P2Y receptor responses are controlled by ectonucleotidases, which usually maintain very low eATP in the pericellular space (29).
Purinergic signaling can be used by the host to activate defense mechanisms, as well as by pathogens to subvert cytoprotective strategies of the eukaryotic cell (30,31).On one hand, eATP, released as a danger signal by injured or stressed cells, plays an important role in the regulation of immune responses, as it triggers purinergic-dependent release of proinflammatory cytokines and chemokines and cell repair processes.On the other hand, toxins such as those that belong to the repeats in toxin-family are known to induce ATP release and the stimulation of P2 receptors that permeabilize the plasma membrane to diffusible ions, and can cause swellingdependent cell lysis (23,(32)(33)(34).
These findings point out that eATP operates as one key regulatory signal in the dynamic pathogen-host cross talk, with the balance of this interaction affecting the infection outcome.

Effect of nucleotide scavengers on AP induction
To test AP induction by S. marcescens, EGFP-LC3-Chinese hamster ovary (CHO) cells were co-incubated with the WT strain for 120 min, after which cells were visualized by confocal microscopy.A S. marcescens shlBA mutant strain lacking ShlA expression was used as a negative control.AP induction by the WT strain was revealed by an EGFP-LC3 green fluorescent punctate pattern, indicative of LC3 recruitment onto nascent autophagosome membranes, as opposed to a homogenous distribution of fluorescence observed when the shlBA mutant strain was used.To determine whether eATP might influence the AP induction response, experiments were run in the absence or presence of an excess of enzymes known to remove eATP (35,36).
Enzymes were added to assay media-containing CHO cells prior to bacterial infection.Results show that apyrase, hexokinase (HK), and Na + ,K + -ATPase, at 20 U/ml, were able to reduce the autophagic response by 63 to 79% (Fig. 1A).Experiments using HK, Na + ,K + -ATPase and apyrase in the absence of cells or bacteria, but in the presence of 60 nM ATP, showed that ATP degradation is complete and rapid, with t 1/2 amounting to 0.1 to 0.5 min (Figs.1B and S1).Inhibition of AP increased nonlinearly as a function of apyrase concentration (Fig. 1C).A hyperbolic function was fitted to data, with K 0.5 = 5.9 U/ml.To test the effect of nucleotides on AP in the absence of toxin, CHO cells were preincubated with 3 μM of ATP, ADP, or UTP before exposure to the shlBA strain.Results show that nucleotides per se did not induce AP.
Overall, our results imply that ATP accumulates in the extracellular medium as a consequence of CHO exposure to ShlA and this eATP triggers AP.The following experiments were designed to test this hypothesis.
Next, we quantified eATP kinetics of CHO cells exposed to noninvasive Escherichia coli W3110 strain transformed with the pES14 plasmid that harbors the shlBA operon.This strain displays 1.8-fold higher hemolytic activity when compared to WT Serratia.E. coli W3110/pES14 at multiplicity of infection (MOI) 10 exhibited a relatively slow [eATP] increase lag phase followed by a fast [eATP] increase to a maximum, indicating activation of ATP release.At the late phase of eATP kinetics, [eATP] decayed indicating eATP hydrolysis by ATPases.Maximum accumulation of eATP with E. coli W3110/pES14 at MOI 10 was 4.5-fold higher than the highest [eATP] obtained using WT Serratia at MOI 100.No changes in [eATP] were observed using the E. coli W3110 strain (Fig. 2A), reinforcing the notion that ShlA is the inducer of ATP efflux from CHO cells.
In the absence of host cells, bacterial [eATP] was stable, so that ShlA was not inducing bacterial ATP release (Fig. S1F).Exposure to melittin (a permeabilizing peptide) caused an acute ATP release from WT Serratia.However, we used 2 × 10 8 bacteria for this experiment, which in the hypothetical presence of CHO cells would represent an effective MOI >2600, that is, 26-fold the MOI employed in experiments shown in Figure 2A.Thus, such bacterial ATP release, even under high bacteriolysis, would not affect eATP kinetics of CHO cells.This means that eATP (Fig. 2) originates entirely from iATP of CHO cells.
Cytotoxicity was assayed in CHO cells challenged by each of the four strains, S. marcescens WT or shlBA and E. coli W3110/pT7 (empty vector) and E. coli W3110/pES14.Cytotoxicity values tested by the thiazolyl blue tetrazolium bromide (MTT) colorimetric assay moderately but continuously increased (Fig. S2A and MTT inset).However, this increase was not related to iATP release, since it was also observed in the strains lacking ShlA expression, which did not induce changes of [eATP].Moreover, as evaluated by a flow cytometry assay, propidium iodide uptake, a marker of cellular membrane damage, showed low values for the exposure of CHO cells to all bacterial strains, thus discarding a lytic component of ATP release (flow cytometry inset, Fig. S2A).
Caco-2 cells were also exposed to either WT S. marcescens (MOI 100) or to E. coli W3110/pES14 (MOI 5 or 10).Sigmoidal patterns of eATP kinetics were observed (Fig. 2B), indicating that Caco-2 cells release ATP when exposed to ShlA.For E. coli W3110/pES14, [eATP] accumulation values increased as a higher MOI was used.eATP accumulation could not be detected when either WT Serratia at MOI 10 or the shlBA strain at MOI 100 were used, probably due to the sensitivity of the methodology used to measure eATP (Fig. 2B).
Results suggest that, in CHO cells exposed to WT Serratia, ATP release is mediated by PNX1 and by exocytosis.

Hydrolysis of eATP by bacteria and CHO cells
Kinetics of eATP shown in Figure 1A not only depends on iATP release (increasing [eATP]) but also on eATP hydrolysis (decreasing [eATP]).Accordingly, experiments were run to assess the capacity of bacteria and CHO cells to degrade eATP (Fig. 4).First, the kinetics of eATP was measured for CHO cells exposed to exogenous ATP.Following ATP addition at three different concentrations, acute [eATP] increases were observed, followed by decay phases (Fig. 4A).By fitting a monoexponential decay function to data, the initial rates of [eATP] decrease at each [ATP] could be determined.This A similar procedure was carried out to assess eATP hydrolysis by Serratia (Fig. 4C).The addition of different concentrations of ATP to the bacterial assay medium caused an initial increase and subsequent decrease in [eATP].After  ShlA-dependent eATP-induced autophagy and egress calculating initial velocities of ATPase activity at each [ATP], ATPase activity (vi) was plotted as a function of [ATP].A linear fit to these data yielded a slope (K ATP-Serratia ) of 0.91 ± 0.22 min −1 mg −1 (Fig. 4D).
The above results imply that, within the micromolar range where eATP accumulates in ShlA-challenged CHO cells, increases in [eATP] concentration can in principle be counteracted by eATP hydrolysis.A quantitative analysis of these results is presented in the Mathematical modeling section below.
Additional experiments were performed to test the maximal capacity of E. coli, Serratia, and CHO cells to hydrolyze nucleotides.The three systems exhibited significant hydrolysis of ATP (Fig. S3), ADP, and AMP (Fig. S3).However, neither nucleotides nor adenosine affected Serratia growth (Fig. S2, B  and C), ruling out the potential effect of a metabolic or (Jv_CHO) was slightly lower than J ATP , and several fold higher than bacterial J v_S.marcescens .F, effects of changing the rate of eATP hydrolysis of CHO cells on eATP kinetics.Predictions of eATP kinetics were made considering the experimentally determined rate constant of eATP hydrolysis ("K ATP " set to 1), a 5-fold increase or a 5-fold decrease of K ATP .CHO, Chinese hamster ovary; eATP, extracellular ATP.
energetic advantage for the bacteria to induce the release of ATP from the eukaryotic cell.

Mathematical modeling of eATP kinetics of CHO cells challenged by Serratia
The observed ShlA-induced eATP kinetics (Fig. 2A) is the result of ATP release by CHO cells and eATP hydrolysis by both ectonucleotidases of CHO cells (Fig. 4A) and periplasmic ATPases of Serratia (Fig. 4C).While eATP kinetics and eATP hydrolysis constitute experimental results, the rate of iATP release can be calculated using a data-driven model.In addition, once the model was fitted to experimental data, it allowed to quantify the contribution of ATP release and eATP degradation to the dynamic regulation of eATP (Fig. 4, E and F).In the model, time-dependent changes in [eATP] were expressed as: ∂½eATP ∂t with J ATP being the rate of iATP release by CHO cells, J v-CHO the eATP hydrolysis by CHO cells, and J v-bact the eATP hydrolysis by Serratia.
The predicted kinetics of the three fluxes are shown in Figure 4E.For CHO cells challenged by Serratia, J ATP (i.e., ATP efflux) displayed a lag phase, followed by a continuous nonlinear increase.J ATP and J v-CHO displayed comparable rates, while J v-bacteria was much lower and therefore unable to affect eATP kinetics.
By comparing J ATP with [iATP] of CHO cells (6.02 ± 1.04 mM), it was possible to calculate the energetic cost of ATP efflux in CHO cells, showing that eATP represented 1 to 2.5% of iATP (Fig. S4A).
The consequences of ectonucleotidase activity of CHO cells can be best observed by considering changes in K ATP , the kinetic constant of the substrate curve shown in Figure 4F.A K ATP value set to 1 (1-fold of the experimental value) allows to model eATP kinetics matching the experimental results.A 5-fold reduction in K ATP , that is, K ATP = 0.2, accelerated a sustained [eATP] increase (Fig. 4F), while a 5fold increase in K ATP (i.e., K ATP = 5) leads to very low values of eATP kinetics.
WT Serratia displays significant hydrolysis of eATP in the low micromolar range (Fig. 4, C and D).However, at MOI = 100, such ATPase activity has no effect on eATP kinetics of CHO cells.This is why simulating eATP kinetics at MOI = 0 or 100 provided similar results (Fig. S4B).For bacterial ATPase activity to affect eATP kinetics, MOI values should have to increase at least two orders of magnitude (Fig. S4B).
In principle, P2Y2 should be activated by low-micromolar eATP to induce AP, no matter whether the nucleotide is produced endogenously from iATP release or exogenously provided.To test this hypothesis, CHO cells were preincubated 20 min with 20 U/ml HK, a treatment capable of efficiently removing ATP (Fig. S1C).Under this condition, 5 μM of the slow degradable analogs ATPγs and AMP-PNP were added together with the Serratia challenge.Results show that, having removed endogenous eATP with HK, both nucleotide analogs increased the AP induction (i.e., reverted the AP inhibition induced by HK) from 52% (HK) to 71% (HK + AMP-PCP) or to 93% (HK + ATPγs).The higher effect of ATPγs versus AMP-PCP agrees well with the high affinity of this analog to P2Y2 (45) (Fig. 5C).
Altogether, results show that in CHO cells AP induction promoted by ShlA is mediated by P2Y2 receptors.In contrast, neither suramin nor apyrase were able to reduce the canonical AP that occurs when CHO cells are challenged by starvation (Fig. S5).
The kinetic behavior of J ATP (Fig. 4E) is similar to the kinetics of AP inhibition provoked by addition of suramin and apyrase (Fig. 6A), as it can be observed when J ATP and AP (in %) were plotted together versus time (Fig. 6B).Apyrase or suramin treatments are less effective as J ATP increases.

Role of integrin α5β1 on P2Y2-dependent induction of AP
Given that P2Y2 has been described to display an RGD motif that is recognized by α5β1 integrin (44), we examined whether the integrin receptor might be involved in the ShlAdependent signaling cascade.For that purpose, before AP was induced by WT Serratia, CHO cells were preincubated with a peptidomimetic integrin antagonist that targets α5β1 integrin with high affinity (46), either in the absence or the presence of 20 U/ml apyrase.
Results showed that, in the presence of WT Serratia, α5β1 blockage reduced AP by 60% (no apyrase) and by 70% (with apyrase) (Fig. 7).These results were similar to those obtained ShlA-dependent eATP-induced autophagy and egress by P2Y receptor blockage or apyrase treatment (Figs. 1 and 5).Lower concentrations (10-100 nM) of the antagonist had reduced potency on AP inhibition (Fig. 7).No response was obtained for the different treatments when the shlBA strain was used (Fig. 7).
Collectively, these results indicate that the α5β1 integrin receptor is involved in the signal transduction cascade that elicits AP in CHO cells exposed to ShlA.
Integrin α5β1 may exist in a dynamic equilibrium between inactive and active states (47).We thus run experiments to  verify whether α5β1 integrin was active when CHO cells were challenged by WT Serratia, both in the absence or the presence of eATP.For that purpose, CHO cells were exposed to 1 mM Mn 2+ , a well-known integrin activating reagent (48), both in the absence or presence of 40 U/ml apyrase.Results show that 1 mM Mn 2+ caused a slight but not significant increase of AP (Fig. S6).This suggests that, in the presence of ShlA, α5β1 is present mostly in an active state.As observed before (Fig. 1A) apyrase highly reduced AP.However, addition of Mn 2+ -in the presence of apyrase-increased AP from 30 to 62%.As expected, when cells were exposed to the shlBA mutant strain, Mn 2+ had no effect.This observation suggests that eATP removal (achieved through apyrase treatment) led to the inactivation of P2Y2, thereby shifting the equilibrium of α5β1 conformations toward inactive state(s).Under this condition, Mn 2+ causes activation, although to a smaller extent than when eATP is present (≈60% versus 100%).

Induction of Serratia egress from the intracellular vacuole
Because we have previously demonstrated that ShlA expression was required to promote the bacterial exocytic, nonlytic egress from the invaded host cell, we examined whether this process could be linked to ShlA-dependent activation of P2Y2.We first analyzed whether the sole extracellular contact of ShlA with cells previously invaded with the shlBA strain could promote these intracellular mutant bacteria (otherwise unable to egress) to be exocyted.The exposure of CHO cells to a noninvading E. coli strain that recombinantly expresses ShlA rescued the shlBA strain in its escape from the invaded cell (Fig. 8A).We also corroborated that bacterial release was not due to a cytotoxic effect on CHO cells (Fig. 8B).
Next, we compared the effect of the inhibitors suramin (P2generic) and AR-C118925XX (P2Y2-selective) on the capacity of WT versus the shlBA mutant strains to egress from CHO cells.Addition of each blocker could prevent WT strain exit, as measured by an increased percentage of intracellular bacteria (Fig. 8C) or by concomitant diminished bacteria released to the culture supernatant (Fig. 8D).Therefore, under the action of either inhibitor, the WT strain mimicked the shlBA mutant strain behavior, while the shlBA mutant inability to exit from CHO cells was not altered by the compounds (Fig. 8, C and D).We also discarded that incubation of CHO cells with 3 μM ATP, ADP, UTP, and adenosine before exposure to the Serratia strains affected bacterial egress (Fig. S7).
Overall, these results show that Serratia ShlA-dependent activation of P2Y2 receptor is an early signaling event that will promote the clearance of Serratia from the invaded host cell and allow bacteria to disseminate extracellularly after intracellular multiplication.

Discussion
Results of this study show that WT S. marcescens induces regulated iATP release from CHO cells.The resulting eATP activates a complex signaling pathway that triggers AP and mediates a postinvasion response, provoking the nonlytic clearance of the pathogen from the host cell.

Induction of noncanonical AP by eATP
Several lines of evidence support the role of eATP in ShlAdependent AP.First, an excess of three unrelated exogenous enzymes capable of degrading eATP blocked the ShlAmediated AP in host cells.Second, WT Serratia, but not the otherwise isogenic shlBA mutant strain, promoted the release of iATP from CHO cells, leading to continuous accumulation of eATP.This response was replicated by exposing cells to a noninvasive E. coli strain transformed with a plasmid that harbors the shlBA operon, which expresses ShlA and its cognate transporter ShlB (12).
Inhibition of PNX1 by low concentrations of CBX (whose binding site is clearly identified in the channel (49), and MFQ, as well as blockage of exocytosis by BFA, strongly suppressed the ShlA-dependent AP response.Because WT Serratia did not induce lysis on CHO cells, these results indicate that ATP release occurs by regulated processes.
In addition to the response on CHO cells, we also observed that ShlA significantly enhanced ATP release from Caco-2 cells, a model of enterocytes and colonic adenocarcinoma (50).This is particularly important considering that S. marcescens is an opportunistic pathogen with an ample range of human host cells, including intestinal epithelial cells (51).

Energetics of eATP regulation
eATP signaling did not impose an energy burden to CHO cells, since eATP accumulation required approximately 2% of total iATP.However, results using the E. coli strain that overexpresses ShlA, and thus displays higher cytotoxic activity than WT Serratia, show that the energy cost for host cells can be higher, potentially contributing to energy depletion of the host.Previous reports showed that ShlA exposure to different eukaryotic cells led to vacuolation due to irreversible ATP depletion, sometimes leading to cytolysis (52).In this study, we show that subtler noncytolytic changes mediated by purinergic signaling are sufficient to elicit the autophagic response.
A low energy requirement by the host to produce eATP is in line with the low [eATP] required to activate ATP-sensitive P2Y receptors (53).Moreover, it also agrees well with the infection cycle of Serratia in which invasion and intracellular proliferation is followed by bacterial egress without compromising the viability of the host cell (20).

P2Y2 signaling
eATP may signal onto one or various P2 receptors of the host cell.By using broad spectrum and subtype-specific P2blockers, we showed that AP in CHO cells was highly blocked by AR-C11895, a highly selective antagonist for P2Y2 (54) that displays a potency in the midnanomolar range (54).This agrees well with an app.K 0.5 (about 16 nM, see Fig. 5B) for AP blockage in CHO cells.Moreover, ShlA-dependent eATP accumulation in CHO cells is congruent with both, the reported affinity for P2Y2, and the above reported low energy requirement of CHO cells.In addition, our results using suramin and AR-C118925 show that, in addition to AP, ShlA-promoted bacterial non-lytic egress also depends on purinergic signaling.Furthermore, we verified a tight coupling between the temporal pattern of AP inhibition and either the suramin or apyrase blockage of P2Y2-dependent ATP release.

Downregulation of eATP-P2Y2 signaling by eATP hydrolysis
Model-dependent fit to experimental results showed that, following exposure of CHO cells to Serratia, [eATP] kinetics depended on both iATP release from host cells, and [eATP] hydrolysis by ecto-ATPase activity of CHO cells.Under the analyzed experimental conditions, and in contrast to other A and B, CHO cells were infected with S. marcescens shlBA strain.After 60 min, extracellular bacteria were killed by gentamicin.At 240 min postincubation (p.i.), antibiotic-free medium and Escherichia coli/pES14 or E. coli/pES15 were added, when indicated.A, at 360 min p.i., CFUs of extracellularly released shlBA strain were determined.B, at 360 min p.i., MTT was added to assess cytotoxicity.Noninvaded cells treated with Triton X-100 were used as a positive control; noninvaded and untreated cells were used as a negative control.Average values ± SD of N = 3 are shown (*p < 0.05; **p < 0.005, ****p < 0.0001).C and D, CHO cells were infected with WT or shlBA strains.At 0 or 240 min p.i., suramin was added.C, at 360 min p.i., intracellular CFU (%) was calculated relative to the inoculum.Average ± SD of N = 3 is shown (*p < 0.05).D, after 240 min p.i., gentamicin was eliminated by replacement of free-antibiotic medium.CFU/ml in supernatant (SN) was determined at 360 min p.i. Data represent means ± SD of N = 3. ** denote p ≤ 0.005 and ***p < 0.001 (two-way ANOVA and Tukey-Kramer multiple comparisons test).CFU, colony-forming unit; CHO, Chinese hamster ovary.
bacteria that are able to hydrolyze substantial eATP even at low MOI (55), rates of eATP hydrolysis by Serratia ATPases were very low and thus unable to alter ShlA-dependent eATP kinetics of CHO cells.

Role of α5β1 integrin
Like other Gαq-coupled P2Y receptors, activation of P2Y2 stimulates the canonical Gαq/phospholipase C/inositol triphosphate signaling axis, leading to release of calcium from intracellular stores.This route appears functional in CHO cells, since micromolar ATP and UTP are able to increase inositol triphosphate and mediate calcium oscillations (56,57).However, the link between Gαq activation and AP in different cell systems is not clear.Although downstream signaling through PI3K/Akt/mTor was shown to inhibit the LC3-I to LC3-II conversion (58), thereby inhibiting the AP response, in CHO cells, we showed that wortmannin, a blocker of PI3K, does not interfere with noncanonical AP induction of WT Serratia (21).
Another signaling route activated by P2Y2 involves the interaction of Arg-Gly-Asp motif within its first extracellular loop with αVβ3/5 and α5β1 integrins, two members of the RGD-recognizing family (44).Because CHO cells exhibit a functional α5β1 (but no β3 integrins), we used a peptidomimetic compound shown to specifically antagonize α5β1 integrin (46).This antagonist dose dependently decreased ShlA-dependent AP up to 40%, implying that P2Y2, activated by eATP, may transactivate α5β1 integrin, leading to AP induction.Alternatively, given that P2Y2 is able to activate five distinct signaling pathways (44), an outside-in signaling modulating α5β1 integrin cannot be discarded.
Because BFA interferes with the central vacuolar system traffic, it may not only alter eATP accumulation and subsequent P2Y2 activation but also the transit of integrins to and from the cell membrane (59), thus affecting transactivation.We have previously shown that intracellular WT Serratia promotes a ShlA-dependent calcium mobilization that leads to dynamic modulation of the cytoskeleton required to induce the exocytosis of bacteria (22).This would be consistent with ShlA promoting P2Y2 transactivation of integrins that results in the Gα12-and Gαo-dependent activation of Rho, Rac, and Cdc42, which give rise to cytoskeletal rearrangements (44).

Reversible injuries to the plasma membrane
It has been shown that injuries to the plasma membrane provoked by bacterial toxins can induce an AP-related process that involves macropinocytosis.This mechanism promotes the maintenance of the plasma membrane integrity and encompasses removal of damaged material and repair of the membrane structure (60).The ShlA-dependent AP induction is reversible and unrelated to the biogenesis of the autophagiclike Serratia-containing vacuoles (21).Therefore, it is tempting to speculate that ShlA-induced P2Y2 activation is required to trigger an AP-related healing process as a shortterm response.In fact, we herein show that the ShlAmediated AP induction does not overlap with the canonical starvation-induced pathway.In principle, brief and reversible injuries occurring in CHO cells exposed to ShlA should lead to pulsed increases of [eATP] (complementing MFQ-and BFAsensitive ATP efflux), which may contribute to the observed residual, MFQ/BFA-tolerant, component of ATP release.
To summarize our main findings a graphical model of our postulated eATP-P2Y2-integrin signaling cascade mechanism modulating ShlA-dependent phenotypes is provided in Figure 9.At least two different transport systems facilitate ShlAdependent iATP release from CHO cells.eATP can then be partially hydrolyzed by ecto-ATPase activity of CHO cells and activate P2Y2 receptors.Next, P2Y2-α5β1 interaction would activate Go/G12 promoting AP and also trigger a nonlytic egress and dissemination of intravacuolar bacteria from the host cell.

Final remarks
Considering that nosocomial infections due to S. marcescens are hard to treat, our results suggests that strategies aimed at interfering with purinergic signaling would be able to suppress the ability of the host cell innate immune response to counter the effects of ShlA and hinder the intracellular traffic of Serratia that leads to egress and spread within the host.Numerous compounds targeting purinergic receptors and/or integrins in clinical contexts are available to treat atherosclerosis, excessive inflammation, cancer, retinal neovascularization, and in age-related macular degeneration (61)(62)(63), so that repurposing these agents or a rational development of new inhibitor molecules can be foreseen as promising therapeutic alternatives to treat Serratia infections.

Experimental procedures
Bacterial strains and plasmids are listed in Table S1.

AP assay
The AP assay was performed as described (20).EGFP-LC3-CHO cells were cultured in 24-well plates until they reached 50% confluence.S. marcescens or E. coli cultures were washed once with PBS, and an appropriate volume was added to each well to reach a MOI of 2. Plates were centrifuged for 10 min at 1000 rpm and incubated for 2 h at 37 C and 5% CO 2 .Then, the cells were washed five times with PBS and fixed with 0.5 ml 3% paraformaldehyde in PBS.When indicated, before exposure to bacteria, CHO cells were pretreated with various blockers of purinergic receptors, or with the nucleotidescavenger enzymes apyrase, HK, and Na + ,K + -ATPase.
Postacquisition image analysis was performed using the ImageJ software (NIH; https://imagej.nih.gov/ij/download.html).At least 200 cells were analyzed for each condition.The results for each experiment are the average of an assay performed in triplicate and independently repeated three times.

Gentamicin protection assay (egress assay)
The gentamicin protection assay was performed as described (21).All infection assays were done at MOI = 10 for CHO cells.Percentage of intracellular colony-forming unit (CFU) was calculated relative to the inoculum.To quantify bacteria in the extracellular medium of invaded cells, gentamicin-containing medium was replaced by free antibiotic medium (22).At indicated time points, the supernatant was recovered and serially diluted.CFUs in supernatant were determined on LB agar plates, and CFU/ml was calculated.The results for each experiment are the average of an assay performed in triplicate and independently repeated at least three times.
Cell viability was determined by the MTT reduction assay and propidium iodide uptake as described before (20).Hemolytic activity assays were performed as previously described (20).eATP and iATP measurements ATP was measured by real-time luminometry as described before (64).Measurements of eATP were carried out with  3) and (ii) eATP can activate with high-affinity PY2 receptors functionally present in CHO cells (4).In turn, activated P2Y2, via its Arg-Gly-Asp motifs in the first extracellular loop, is able to transactivate α5β1 integrin (5).Alternatively (not visualized), outside-in signaling of P2Y2 might indirectly modulate α5β1 integrin.On speculative basis, the P2Y2-α5β1 interaction is necessary for P2Y2 to activate Go/ G12 signaling routes, thus inducing two distinct phenotypes: an early AP response (7) and postinvasion cytoskeletal rearrangements allowing Serratia, replicating in an intracellular vacuole (SecV) (8), to be exocyted (9).A potential endocytotic entry of Serratia into CHO cells is illustrated.CHO, Chinese hamster ovary; eATP, extracellular ATP; iATP, intracellular ATP.
CHO and Caco-2 cells alone, or coincubated with bacteria.Aliquots containing 75,000 cells (with or without bacteria) were incubated in 100 μl of PBS medium.Results were expressed as [eATP] at every time point of a kinetic curve (i.e., eATP kinetics), with [eATP] expressed as μM/(mg protein).
Increases in [eATP] were evaluated as the difference between [eATP] at a fixed time point post stimulus and the basal [eATP] and are indicated as ΔATP 120 (i.e., 120 min post stimulus).
The (iATP) content of CHO cells and S. marcescens was estimated in real-time measurements as previously described (33).ATP values were expressed as iATP concentration.

Hydrolysis of extracellular nucleotides
Maximal hydrolysis rates of ATP, ADP, and AMP were determined by the malachite green method as described (65).Cells (30,000-60,000/300 μl) were exposed to 500 μM of ATP, ADP, and AMP.The content of Pi was determined at different times.For bacteria, 100 μl aliquots of the bacterial suspension (10 9 /ml) were withdrawn as duplicate at different times.After fitting an exponential function to data, initial velocities were calculated and expressed as app Vmax in nmol Pi/μg min.The rate of eATP hydrolysis of CHO cells and S. marcescens at low-ATP concentrations was determined by real-time luminometry (33).ATPase activity was expressed as (vi) in nmol/mg min.

Data analysis
Statistical analysis was performed using one-way ANOVA or two-way ANOVA and Tukey-Kramer Multiple Comparisons test or t test as appropriate with an overall significance = 0.05.Asterisks in the plots denote the values among the treatment groups in which a statistically significant difference was determined.
For experimental results of Figure 4B (AP versus [AR-C118925XX]), a hyperbolic function of the form: y ¼ a:b b þ x þ yo was fitted to data, with "y" being AP (%), "yo" the asymptotic maximal value of AP inhibition, and "b" (K 0.5 ) representing the concentration of AR-C118925XX at which a semimaximal inhibition of AP is obtained.
Details of the mathematical model used to quantify the dynamics of eATP regulation are given in Figure 6D.

Figure 1 .
Figure 1.Inhibition of ShlA-dependent AP of CHO cells caused by nucleotide scavengers.A, CHO-EGFP-LC3 were incubated with 20 U/ml of apyrase, hexokinase, or Na + ,K + ATPase.After 15 min, cells were coincubated with WT or shlBA Serratia marcescens strains.At 120 min coincubation (c.i.), cells were visualized by confocal microscopy (right panels) and ShlA-dependent AP was calculated relative to AP in the absence of enzymes.The scale bars represent 10 μm.B, eATP degradation in the presence of 20 U/ml apyrase, in the absence of cells.At 2.2 min, 60 nM ATP was added, and the eATP kinetics was quantified.[eATP] was estimated by light production a.u (arbitrary units).Similar experiments were run (Fig. S1) using hexokinase and Na + ,K + -ATPase, which allowed to derive half-time (t 1/2 ) values of eATP decay due to the activities of the three enzymes (inset).C, effect of apyrase on AP inhibition.A hyperbolic function (continuous line) was fitted to data.Data represent mean values ± SD of N = 4. **** denote p ≤ 0.0001, two-way ANOVA, and Tukey-Kramer multiple comparisons test.AP, autophagy; CHO, Chinese hamster ovary; eATP, extracellular ATP.

Figure 3 .
Figure 3.Effect of inhibitors of iATP release on eATP and autophagy.The time course of [eATP] from CHO cells was quantified as in Figure 2. Then (A) increases in [eATP] were evaluated as the difference between [eATP] at 120 min poststimulus and the basal [eATP] and are indicated as ΔATP 120 .Cells were preincubated 10 min with 10 μM carbenoxolone (CBX) or 100 nM mefloquine (MFQ), or 3 h with 0.1 μM brefeldin A (BFA).Five minutes following pretreatment, CHO cells were exposed to Serratia marcescens WT strain (MOI = 100).Control experiments were run in the absence of inhibitors.B, similar treatment as in (A) was applied to determine relative values of autophagy (AP, in %), quantified as described in Figure 1.Values of ΔATP 120 taken from (A) are shown for a comparative purpose.CHO, Chinese hamster ovary; eATP, extracellular ATP; iATP, intracellular ATP; MOI, multiplicity of infection.

Figure 4 .
Figure 4. Rates of eATP hydrolysis by CHO cells and Serratia marcescens and model predictions.A and B, CHO cells.C and D, WT S. marcescens.A and C, show eATP kinetics in the absence and presence of increasing [ATP].Arrows show time of ATP addition.For (B) (CHO cells) and (D) (WT S. marcescens), each point ± SD of the graphs was calculated from initial velocities (vi) of eATP decay kinetics measured in (A) and (C), respectively.Linear functions were fitted to data.Data represent mean values ± SD of N = 4. E and F, results of a model showing aspects of eATP hydrolysis for CHO cells exposed to WT S. marcescens.Details of the model in Fig. S8.E, modeled fluxes of ATP.J ATP = ATP efflux from CHO cells, Jv-CHO = rate of eATP hydrolysis by ectonucleotidases of CHO cells, J v_S.marcescens = eATP hydrolysis by ATPases of S. marcescens.(Jv_CHO) was slightly lower than J ATP , and several fold higher than bacterial J v_S.marcescens .F, effects of changing the rate of eATP hydrolysis of CHO cells on eATP kinetics.Predictions of eATP kinetics were made considering the experimentally determined rate constant of eATP hydrolysis ("K ATP " set to 1), a 5-fold increase or a 5-fold decrease of K ATP .CHO, Chinese hamster ovary; eATP, extracellular ATP.

Figure 6 .
Figure 6.Kinetics of autophagy inhibition in the presence of apyrase and suramin.A, inhibition of ShlA-dependent autophagy of CHO cells in the presence of suramin 100 μM or apyrase 20 U/ml added at different times coincubation (c.i.).CHO-EGFP-L3 cells were exposed to WT Serratia marcescens.At 120 min c.i., cells were fixed and visualized by confocal laser microscopy.Time-dependent inhibition of autophagy was assessed.Results are expressed as percentage inhibition relative to the autophagy caused by ShlA in the absence of treatments.B, representative images of CHO-EGFP-LC3 exposed to suramin.The scale bars represent 10 μm.Data represent mean values of N = 4 independent experiments.CHO, Chinese hamster ovary.

Figure 7 .
Figure 7. ShlA-dependent autophagy of CHO cells in the presence of an integrin α5β1 antagonist.CHO-EGFP-LC3 were preincubated 60 min with various concentrations of an integrin α5β1 antagonist (PM-α5β1) in the absence and presence of 20 U/ml apyrase.As a negative control, cells were preincubated in the presence of 1 μM of a αvβ3 integrin antagonist (PM-αvβ3).Following pretreatment, cells were coincubated with WT or shlBA S. marcescens strains.At 120 min c.i., cells were visualized by confocal microscopy.Results are expressed as the percentage of ShlA-dependent autophagy of CHO cells.CHO, Chinese hamster ovary; c.i., coincubation.

Figure 8 .
Figure 8. Purinergic dependent escape of Serratia marcescens from CHO cells.A and B, CHO cells were infected with S. marcescens shlBA strain.After 60 min, extracellular bacteria were killed by gentamicin.At 240 min postincubation (p.i.), antibiotic-free medium and Escherichia coli/pES14 or E. coli/pES15 were added, when indicated.A, at 360 min p.i., CFUs of extracellularly released shlBA strain were determined.B, at 360 min p.i., MTT was added to assess cytotoxicity.Noninvaded cells treated with Triton X-100 were used as a positive control; noninvaded and untreated cells were used as a negative control.Average values ± SD of N = 3 are shown (*p < 0.05; **p < 0.005, ****p < 0.0001).C and D, CHO cells were infected with WT or shlBA strains.At 0 or 240 min p.i., suramin was added.C, at 360 min p.i., intracellular CFU (%) was calculated relative to the inoculum.Average ± SD of N = 3 is shown (*p < 0.05).D, after 240 min p.i., gentamicin was eliminated by replacement of free-antibiotic medium.CFU/ml in supernatant (SN) was determined at 360 min p.i. Data represent means ± SD of N = 3. ** denote p ≤ 0.005 and ***p < 0.001 (two-way ANOVA and Tukey-Kramer multiple comparisons test).CFU, colony-forming unit; CHO, Chinese hamster ovary.