Staphylococcal leukotoxins trigger free intracellular Ca2+ rise in neurones, signalling through acidic stores and activation of store-operated channels

Headache, muscle aches and chest pain of mild to medium intensity are among the most common clinical symptoms in moderate Staphylococcus aureus infections, with severe infections usually associated with worsening pain symptoms. These nociceptive responses of the body raise the question of how bacterial infection impinges on the nervous system. Does S. aureus, or its released virulence factors, act directly on neurones? To address this issue, we evaluated the potential effects on neurones of certain bi-component leukotoxins, which are virulent factors released by the bacterium. The activity of four different leukotoxins was verified by measuring the release of glutamate from rat cerebellar granular neurones. The bi-component γ-haemolysin HlgC/HlgB was the most potent leukotoxin, initiating transient rises in intracellular Ca2+ concentration in cerebellar neurones and in primary sensory neurones from dorsal root ganglia, as probed with the Fura-2 Ca2+ indicator dye. Using pharmacological antagonists of receptors and Ca2+ channels, the variations in intracellular Ca2+ concentration were found independent of the activation of voltage-operatedCa2+ channels or glutamate receptors. Drugs targeting Sarco-Endoplasmic Reticulum Ca2+-ATPase (SERCA) or H+-ATPase and antagonists of the store-operated Ca2+ entry complex blunted, or significantly reduced, the leukotoxin-induced elevation in intracellular Ca2+. Moreover, activation of the ADP-ribosyl cyclase CD38 was also required to initiate the release of Ca2+ from acidic stores. These findings suggest that, prior to forming a pore at the plasma membrane, leukotoxin HlgC/HlgB triggers a multistep process which initiates the release of Ca2+ from lysosomes, modifies the steady-state level of reticular Ca2+ stores and finally activates the Store-Operated Calcium Entry complex.


Introduction
Staphylococcus aureus is a common host in human flora that asymptomatically colonizes one in three healthy individuals; children having higher persistent carriage rates than adults (Wertheim et al., 2005). However, this bacterium can be a serious pathogen, and is now the second cause of hospital and community-acquired infections worldwide (David and Daum, 2010;DeLeo et al., 2010;David et al., 2011). The severity and locations of S. aureus infections display broad variation. Contamination typically progresses from a local infection (furuncle, cellulitis, wound infection) to systemic propagation (bacteraemia) and metastatic infections (pneumonia, endocarditis, osteomyelitis, septic arthritis). The emergence and spread of community-acquired, drug-resistant (CA-MRSA) and extremely virulent strains reinforces the morbidity of S. aureus infections (Chambers and DeLeo, 2009;David and Daum, 2010). Pathogenesis of the bacterium relies on an arsenal of virulence-associated factors, some of which specifically produce toxin-induced diseases (e.g. bullous impetigo, toxic shock syndrome, staphylococcal scaled skin syndrome and food-borne gastroenteritis) (Archer, 1998;Lowy, 1998).
Major advances have been achieved over the last decade in the understanding of the structure and mechanisms of pore formation by two-component toxins from S. aureus and from other origins (Menestrina et al., 2003;Gonzalez et al., 2008). In contrast, much less is known regarding the overall cell targets and responses to S. aureus leukotoxins. The increase in free intracellular calcium or the loss of cellular potassium are examples of early cellular reactions to the presence of pore-forming toxins, an ion imbalance that may lead to the activation of a variety of signalling cascades (Woodin and Wieneke, 1963;Staali et al., 1998;Kloft et al., 2009;Kao et al., 2011).
In the present study, the question of the sensitivity of neurones to bi-component leukotoxins of S. aureus is addressed with the aim of deciphering the mechanisms of cell response. Cell cultures from rat cerebellar cortex and primary sensory neurones from dorsal root ganglia were used to demonstrate the neurotoxic activity of PVL, a-toxin and g-haemolysins. The four virulence-associated factors maintained neuronal integrity while triggering a [Ca 2+ ]i increase in cerebellar neurones, which was followed by the release of glutamate. The g-leukotoxin HlgC/ HlgB was found to be the most potent on both types of cultured neurones. The leukotoxin-induced rises in free intracellular ([Ca 2+ ]i) were initiated by a discharge of Ca 2+ from acidic stores and followed by a Ca 2+ -induced Ca 2+ release from the reticulum. These effects were further amplified by the activation of the store-operated Ca 2+ entry complex. The initial signal linking leukotoxin binding to the acidic stores was due to the activation of the ADP ribosyl cyclase CD38. This cellular response to the presence of the leukotoxin in neuronal cells preceded the formation of toxin pores into the plasma membrane.

Results
Staphylococcal leukotoxins trigger glutamate release from cerebellar granular neurones Cerebellar granular neurones have been shown to act in response to the presence of certain pore-forming toxins (Lonchamp et al., 2010). Thus, to examine whether neurones sense and react to S. aureus leukotoxins, glutamate release by cerebellar granular neurones was quantified after a 10 min exposure to the toxins. Doseresponse curves for a-toxin, leukotoxins HlgC/HlgB and HlgA/HlgB and the PVL are shown in Fig. 1A. HlgC/HlgB was the most potent of the four leukotoxins, reaching a half-maximal effect at a concentration of 3 nM and maximum glutamate release at 0.1 mM. This maximum was 2.4 times higher than the total release generated by a 5 s depolarization pulse of 60 mM KCl, but only 35% of the total-cell glutamate content, as assessed using a 10 min osmotic shock in 10 mM Tris-HCl, pH: 7.4 ( Fig. 1B). Alpha-toxin (0.1 mM), leukotoxin HlgA/HlgB (0.1 mM) and PVL (0.1 mM) were less effective than HlgC/ HlgB, releasing equivalent amounts of glutamate as that generated by the 5 s 60 mM KCl depolarization pulse (Fig. 1A). The release due to leukotoxin HlgC/HlgB was dependent on an increase in free [Ca 2+ ]i, since preincubation of neurones with the intracellular calcium chelator BAPTA-AM (10 mM) (30 min loading followed by 20 min hydrolysis) strongly reduced the amount of glutamate released into the medium (Fig. 1B). The HlgC or HlgB subunits separately failed to induce glutamate release from granular neurones (Fig. S1a).
To determine whether HlgC/HlgB-treated neurones undergo lysis, the amount of glucose-6-phosphate dehydrogenase (G6PDH) enzymatic activity released into the medium was quantified. An example of G6PDH activity measured after incubation at increasing concentrations of HlgC/HlgB leukotoxin is illustrated in Fig. 1C. Significant enzymatic activity was detected only after 20 min incubation with 0.1 mM leukotoxin, suggesting that high toxin concentration treatment may have been fatal for some cells. However, assessment of the number of pyknotic nuclei in cultures labelled with Hoechst-33258 revealed no significant difference between control (three independent cultures, > 400 nuclei, 6.5% pyknotic) and 6 nM leukotoxin-treated cells after a 15 min incubation period (three independent cultures, > 400 nuclei, 5.8% pyknotic). Membrane permeation of ethidium bromide has been proposed for monitoring pore formation at the cell membrane due to leukotoxin (Staali et al., 1998). Time-lapse video microscopy was used to record the infiltration of ethidium bromide in neurones incubated in the presence of 8 nM leukotoxin HlgC/HlgB-Alexa-488. Visual inspection of neurones incubated 30 min in the presence of toxin revealed that ethidium bromide-related fluorescence remained low, constant and consistently extracellular (Fig. S2). These results together with the reduction in glutamate release due to the Ca 2+ chelator BAPTA favour the notion that leukotoxin-induced glutamate release is triggered by a rise in free [Ca 2+ ]i, independently of the formation of a pore.
The typical reaction of a neurone to the presence of leukotoxin HlgC/HlgB was a progressive elevation in free [Ca 2+ ]i which led to a transient peak and returned to a new [Ca 2+ ]i, resting value ( Fig. 2A and B). In a same culture plate, the effect of the leukotoxin on different granular neurones occurred at various unrelated moments. However, the response of DRG sensory neurones was both fast and synchronous for all responding cells of the same plate. For comparison purposes, the response of each neurone was characterized by the peak value of [Ca 2+ ]i (amplitude) and the time needed by the cell to reach the [Ca 2+ ]i peak after the addition of the leukotoxin (latency). These values are shown in box plots with the mean, the median and percentiles.
In granular neurones, the amplitude of the [Ca 2+ ]i peak was dependent on leukotoxin concentration, as illustrated in examples in Fig. 3A-C. Moreover, the consolidation of results from 44 independent experiments confirmed that mean [Ca 2+ ]i peak amplitude values increased, while the latency of the response decreased with increasing concentrations of leukotoxin, as shown in Table 1   A. Cultured cells were challenged for 10 min with the indicated concentrations of leukotoxin HlgA/HlgB ( ), leukotoxin HlgC/HlgB ( ), a-toxin ( ) or PVL ( ). Leukotoxin HlgC/HlgB was the most active, showing a half-maximum concentration effect at nanomolar concentrations. The other three toxins triggered glutamate release at higher concentrations only; the glutamate measured in these instances was similar to the amount measured after a 10 s pulse of 60 mM KCl depolarization. B. Prior to addition of the toxin, cells were loaded with 10 mM of the membrane-permeable intracellular calcium chelator BAPTA-AM ( ), which led to a significant decrease in the amount of glutamate release. Control leukotoxin HlgC/HlgB ( ). The upper reference value (6.51 mM, blue dotted line) stands for the total-cell glutamate content. C. Glucose-6-phosphate dehydrogenase enzymatic activity measured in culture media after a 10 min incubation of neurones in the presence of leukotoxin HlgC/HlgB. Only the highest concentration (0.1 mM) demonstrated a significant amount of released enzymatic activity. Results in (B) (glutamate) and (C) (G6PDH) are from of the same experiment. The brown dotted line represents glutamate (A and B) or G6PDH activity (C) measured in the bathing media of control cells. The green dotted line in (A) and (B) corresponds to glutamate released by a 10 s depolarization of neurones in 60 mM KCl. The blue dotted line in (B) and (C) stands for total glutamate (B) or total G6PDH (C) content measured after a 10 min osmotic shock of neurones in 10 mM Tri-HCl buffer (pH 7.4). and E. However, the dispersion of both parameters was high among neurones within a same plate, which suggests that a multistep process may be necessary to disrupt [Ca 2+ ]i homeostasis and that its activation may depend on a particular cell's status.

The increase in free [Ca 2+ ]i is independent of voltage-operated Ca 2+ channels or glutamate receptor activity
Activation of any of the various types of voltage-operated Ca 2+ channels (VOCC) or glutamate receptors may lead to rise in [Ca 2+ ]i in granular neurones (Randall and Tsien, 1995;Brickley et al., 2001). To investigate whether leukotoxin HlgC/HlgB activates a Ca 2+ -permeable channel or receptor, VOCC were antagonized with either 5 mM nifedipine (L-Type channels), 0.1 mM w-GVI-A (N-Type channels) or 0.1 mM w-AgaTK (P/Q-Type channels) whereas ionotropic glutamate receptors were antagonized with 10 mM CNQX (AMPA receptors) or 10 mM D-AP5 (NMDA receptors) while recording [Ca 2+ ]i variations induced by leukotoxin HlgC/HlgB. Amplitude and latency of [Ca 2+ ]i variations are presented in Fig. 4A and B and in Table 2 (mean values). The presence of these drugs either alone or in various combinations, prior to addition of the leukotoxin, did not considerably alter the amplitude of the [Ca 2+ ]i peak (Fig. 4A). However, blockade of all VOCC significantly shortened the latency of the leukotoxin effect by 45% (P < 0.05), effect observed also when Na + channels and glutamate receptors were additionally blocked (Fig. 4B, Table 2). When the P/Q-type Ca 2+ channel antagonist w-AgaTk was removed from the cocktail, the latency of the response was four times longer (853 Ϯ 14 s versus 210 Ϯ 20 s, P < 0.05). The selective blockade of voltage-gated Na + channels (0.3 mM TTX) doubled the latency of the response (436 Ϯ 8 s versus 210 Ϯ 20 s, P < 0.05). Altogether, these results rule out both VOCC and glutamate receptors as direct passageways for extracellular Ca 2+ influx induced by leukotoxin. However, given the impact of the blockade of these channels on the kinetics of the leukotoxin effect, a role for resting [Ca 2+ ]i or membrane potential cannot be excluded.
To ascertain the contribution of extracellular Ca 2+ , neurones were equilibrated in buffers of various extracellular   (Churchill et al., 2002). Application of GPN (200 mM, 10 min) essentially blunted the leukotoxin effect; only 11% of neurones . None of the differences in mean values for peak amplitude were statistically significant when examined by the Kruskal-Wallis analysis of variance. All the differences in mean latency values were statistically significant (P < 0.001) with respect to the control value (box a); pairwise multiple comparison showed a non-significant difference among values for b and e. Table 2. Mean values and statistical significance of recordings with blunted voltage-operated Ca 2+ channels (VOCC) and voltage-gated Na + channels (VGNC). Mean values and statistical significance of peak amplitudes and latencies of the experiment shown in Fig. 4. The pairwise comparison between the means of the peak amplitude did not reveal significant statistical differences, whereas mean latencies were significantly different, but for experiments b and e only.  (Bird et al., 2008;Varnai et al., 2009). A highly effective neutralization was observed when 100 mM Gd 3+ was added a few minutes prior to the leukotoxin (Fig. 6A). At lower concentration, Gd 3+ (50 mM) reduced the [Ca 2+ ]i amplitude by 44% (291 Ϯ 20 nM versus 519 Ϯ 20 nM; n > 50, P < 0.001) and slightly affected the time of the response, but totally abolished the second Ca 2+ rise present in 96% of control cells challenged with 4 nM leukotoxin (not shown). The SOCE antagonist econazole (5 mM), totally prevented the toxin-induced rise in [Ca 2+ ]i (Fig. 6A).
In addition, incubation of cells for 15 min in 100 mM 2-aminoethoxydiphenylborane (2-APB), a cell-permeable modulator of inositol (1,4,5)-P3-induced Ca 2+ release targeting SOCE, decreased the peak amplitude by 40% (203 Ϯ 18 nM versus 335 Ϯ 17 nM; n > 20, P < 0.001) and increased the latency of the response by more than 10-fold (1149 Ϯ 47 s versus 100 Ϯ 1 s; n > 20, P < 0.001) (Fig. 6B). Cells maintained for 15 min in the presence of 90 mM dentrolene, which also blocks ryanodine and Ins(1,4,5)-P3 receptors, showed a 57% reduction in amplitude (144 Ϯ 4 nM versus 335 Ϯ 17 nM; n > 40, P < 0.001) and a sixfold delay in latency (608 Ϯ 25 s versus 100 Ϯ 1 s; n > 40, P < 0.001) (Fig. 6B). The pyrazole derivative YM-58483 has been reported as an inhibitor of the Ca 2+ release-activated Ca 2+ (CRAC) channels (Ishikawa et al., 2003). Pre-incubation of neurones in 10 mM YM-58483 significantly modified the leukotoxin effect by reducing the amplitude of the Ca 2+ peak by 70%  (Fig. 6C). Following the Ca 2+ depletion, the translocation of STIM1 molecules to endoplasmic reticulum-plasma membrane junctions activates the Orai1-formed CRAC channels; simultaneously the formation of STIM1 puncta can be observed using fluorescence labelling methods (Walsh et al., 2009). The consequences of leukotoxin HlgC/HlgB activity on the redistribution of interacting partners of the SOCE complex were investigated through immunolabelling of neurones with antibodies raised against STIM1 and Orai1. Figure 7 illustrates the differences in labelling observed for STIM1 in control cells relative to cells incubated for 10 min with 4 nM leukotoxin. In treated cells, labelling became partially punctuated suggesting that a redistribution of STIM1 is also an effect of leukotoxin HlgC/HlgB. Taken together, the above results from pharmacological manipulation as well as immunolabelling allow to decompose [Ca 2+ ]i movements induced by the leukotoxin as a succession of events beginning by the release of acidic Ca 2+ stores and resulting in the stimulation of the SOCE complex.

The HlgC/HlgB-induced rise in [Ca 2+ ]i is initially dependent on ADP-ribosyl cyclase CD38 signalling
We next focused on the early signalling which could be activated by the leukotoxin in the plasma membrane and induce the release of Ca 2+ from acidic stores. Lysosomes release Ca 2+ through two-pore channels that are opened by nicotinic acid adenine dinucleotide phosphate (NAADP), an intracellular messenger produced by the ADP-ribosyl cyclase CD38 (Zhu et al., 2009;Cosker et al., 2010). Although efficient antagonists of CD38 activity are not currently available, it is known that b-NAD rapidly induces the internalization of the enzyme and reduces its activity (Zocchi et al., 1999). Hence, we assessed the potential effect of preventing granular neurones from being activated by the leukotoxin by incubating the latter for 10 min in the presence of 5 mM b-NAD prior to the addition of the toxin. In four independent experiments, only 26% of recorded cells showed a transient rise in [Ca 2+ ]i, along with a significant reduction in mean peak amplitude (191 Ϯ 14 nM versus 444 Ϯ 18 nM; n = 23, P < 0.001) and a delayed latency (1147 Ϯ 31 s versus 537 Ϯ 35 s; n = 23, P < 0.001) (Fig. 8A). Similar effects were observed using 2.5 mM NADP or 1 mM NAADP added 10 min prior to the toxin (Fig. 8A). NAADP likely induces a desensitization of the lysosomal two-pore channel (Zhu et al., 2009) that may prevent leukotoxin signalling. Challenge of the granular neurones by 5 mM b-NAD or 2.5 mM NADP did not produce [Ca 2+ ]i peaks, but induced a slight increase in [Ca 2+ ]i level.
Granular neurones incubated in Alexa 488-tagged HlgC/HlgB leukotoxin (4 nM) were observed by confocal microscopy after fixation and paired labelling the anti-CD38 antibody or Alexa 594-conjugated Cholera toxin-B subunit ( Fig. 8C and D). The anti-CD38 labelling was observed at the edge of the cells, showing a dotted aspect that partially appeared superimposed with leukotoxin labelling. However, the fluorescence of Alexa 488-HlgC/ HlgB leukotoxin appears located preferentially in the cytosol. Labelling with the Alexa 594-Cholera toxin, which binds membrane gangliosides, surrounded the cells and was clearly segregated from the Alexa 488-leukotoxin fluorescence. These results suggest that the leukotoxin may be partially trapped in intracellular organelles together with CD38.
Taken together, the results presented here can be better understood if confronted to the mechanism proposed for the endolysosomal Ca 2+ signalling in a variety of cells (Morgan et al., 2011). It is suggested that NAADP initiates a multistep process by releasing Ca 2+ from acidic stores, which triggers the further release of reticular Ca 2+ and finally causes the activation of CRAC channels. The HlgC/HlgB leukotoxin could initiate the same signalling pathway, as we suggest through the schematic representation of Fig. 9.

S. aureus leukotoxins are a critical threat to the nervous system
Staphylococcus aureus has been shown to efficiently invade human brain microvascular endothelial cells, which results in tissue damage, brain abscesses and meningeal inflammation (Pedersen et al., 2006;Sheen et al., 2010). We carried out this work in cerebellar granular neurones to assess whether leukotoxins may cause damage to nervous tissue. Rat granular neurones have Fig. 7. Distribution of STIM1 and Orai1 molecules of the SOCE complex before and after incubation of neuronal cells with 4 nM HlgC/HlgB-Alexa 488. Confocal acquisition images of cerebellar neurones labelled with the anti-Orai1goat polyclonal antibody D15 (A1, B1) and the anti-Stim1 mouse monoclonal antibody A-8 (A2, B2). Panels A3 and B3 correspond to merged images with Hoechst 33258-labelled nuclei. Cells in row A were fixed prior to incubation for 10 min in the presence of 4 nM leukotoxin HlgC/HlgB; cells in row B were fixed after incubation with the toxin.
been used for cellular studies on bacterial toxins while glutamate release assays have aided in investigating their neurotoxicity (Foran et al., 2003;Lonchamp et al., 2010). The use of these cells allowed us to establish that different S. aureus leukotoxins can represent a serious threat to neurones by inducing a rise in free intracellular Ca 2+ and, as a consequence, the release of glutamate. The increase in free [Ca 2+ ]i was also measured in DRG neurones, highlighting the fact that the activity of leukotoxins is not restricted to a single class of neurones. The g-haemolysin HlgC-HlgB was found to be the most potent, both in cerebellar and in DRG neurones, inducing important cellular effects while preserving the integrity of the plasma membrane. The pharmacological manipulations, which can prevent free [Ca 2+ ]i movements through the disruption of internal stores as well as acting on the plasma membrane, together with the lack of permeability to ethidium bromide, parallel to the [Ca 2+ ]i increase, strongly supports the notion that the formation of a pore is not the first cellular effect induced by leukotoxins in neurones. Cellular differences between neurones and neutrophils may explain an expanded delay between intracellular Ca 2+ rise and pore formation as it has been observed in neutrophils through the introduction of mutations in PVL (Baba Moussa et al., 1999). Moreover, PVL can confer an enhanced bactericidal capacity to human neutrophils at very low concentrations, independently of the formation of a pore (Graves et al., 2012). The activity of the HlgC/HlgB leukotoxin is concentration dependent, which may suggest that minute amounts of toxin can be sensed by neurones and induce intracellular signalling. A neuronal early response to leukotoxin could have the positive outcome of initiating cellular protection (Aroian and van der Goot, 2007), conceivably in association with noxious stimuli.
Neurones preferentially use VOCC and ionotropic neurotransmitter receptors as Ca 2+ influx pathways, while for neutrophils the SOCE complex is the main Ca 2+ signalling system. This difference could obscure the fact that both cells types are sensitive to the leukotoxin. Interestingly, the Ca 2+ activated movement shown herein demonstrates a reasonable link between internal Ca 2+ stores and membrane signalling proteins, through pathways which are otherwise observed separately in neurones. The ADP-ribosyl cyclase CD38 has been detected in rat brain (Yamada et al., 1997;Ceni et al., 2003), where the signalling molecule NAADP has been proposed as a stimulator of neurite outgrowth (Brailoiu et al., 2005). The release of Ca 2+ from either lysosomes or Ryanodinesensitive pre-synaptic stores has been shown to enhance the efficiency of neurotransmitter release (Collin et al., 2005;McGuinness et al., 2007). In cortical and hypocampal neurones, differential regulatory roles of intracellular Ca 2+ levels have been observed for STIM1 and STIM2 (Gruszczynska-Biegala et al., 2011), and the depletion of Orai1 impairs the rhythmic firing flight motoneurones in drosophila (Venkiteswaran and Hasan, 2009). A further question will be to ascertain whether endogenous activator(s) would be able to stimulate this 'leukotoxin signalling multistep process' in the nervous system. The future identification of the molecular complex that binds the leukotoxins will help to find the answer, as it will contribute to assess the consequences of the Ca 2+ imbalance caused by leukotoxins in neurones. It is of interest to note that in mild S. aureus pathologies, such as chronic rhinosinusitis, associated headaches and facial pain/pressure negatively affect quality of life and impair daily activities (Singhal et al., 2011).

A multistep Ca 2+ signalling mechanism that could be shared by other toxins
An increase in free intracellular Ca 2+ has already been described as an early response for the activity of different pore forming toxins. Intracellular Ca 2+ movements are regarded as a prerequisite for the pre-pore to pore transition of membrane bound bi-component S. aureus leukotoxins (Woodin and Wieneke, 1963;Staali et al., 1998). The a-haemolysin from Escherichia coli has been shown to induce [Ca 2+ ]i oscillations in renal epithelial cells that are sensitive to L-type Ca 2+ channel blockers (Uhlen et al.,Fig. 9. Schematic representation of the multistep signalling mechanism triggered by g-leukotoxin HlgC/HlgB on neuronal cells. Upon binding (1) to a receptor complex, likely associated with lipid rafts, the bi-component leukotoxin triggers the formation of NAADP by ADP-ribosyl cyclase CD38. NAADP, acting on the two-pore Ca 2+ channels of lysosomes (2) elicits the release of Ca2+ into the cytosol. Free [Ca 2+ ]i can further be increased (3) through a Ca 2+ -induced Ca 2+ release mechanism from reticular stores. Finally, reticular Ca 2+ depletion activates the Store-Operated Ca 2+ Entry complex formed by STIM1 and Orai1 (4). 2000), although the contribution of the L-type channel was subsequently challenged (Koschinski et al., 2006). In the present study, the pharmacological neutralization of voltage-operated Ca 2+ channels, including L-type channels, as well as glutamate receptors revealed that these channels are not a primary pathway for the S. aureus leukotoxin-induced rise in [Ca 2+ ]i. However, the latency to reach the [Ca 2+ ]i peak was affected by combinations of drugs known to be active on voltage-operated channels, which may suggest an influence of either the resting [Ca 2+ ]i, or the membrane potential, on the kinetics of leukotoxin activity. The vacuolating cytotoxin A (VacA) from Helicobacter pylori induces the oscillation of free cytosolic Ca 2+ through the mobilization of internal stores (De Bernard et al., 2005). The leukotoxin HlgC-HlgB shares the feature of mobilizing internal stores. Its activity was significantly inhibited when reticular Ca 2+ stores were depleted. Even though the total reticular Ca 2+ load remains low, and free cytosolic [Ca 2+ ]i slightly changes upon blocking the SERCA pump with thapsigargin (Pinilla et al., 2005), the reticulum is an important link for the response process activated by the toxin. A drastic inhibition was observed when H-ATPase, which contributes to the refilling of acidic Ca 2+ stores, was blocked by Bafilomycin A, or upon disruption of lysosomes by GPN. Therefore, we propose that an initial release of Ca 2+ from the acidic stores may be followed by a Ca 2+ -induced Ca 2+ release from the endoplasmic reticulum. The Ca 2+ depletion of the endoplasmic reticulum, sensed by the stromal interacting molecule STIM1, subsequently leads to the assembly of Orai1 molecules to form the store-operated CRAC channel . Noteworthy, listeriolysin O, produced by Listeria monocytogenes, mobilizes Ca 2+ from the extracellular milieu as well as from various types of intracellular stores (Gekara et al., 2007) whereas Streptolysin O, from Streptococcus pyogenes, induces long-lasting intracellular Ca 2+ oscillations through the activation of the SOCE complex, which are modulated by the level of expression of STIM1 and Orai1 (Usmani et al., 2012). Another leukotoxin, secreted by Actinobacillus actinomycetemcomitans, induces free [Ca 2+ ]i movements if the molecules needed for its activity are previously clustered into lipid rafts (Fong et al., 2006). In the present study, the inhibition of S. aureus leukotoxin activity through LY294002 inactivation of PI3K and PI4K may suggest a similar prerequisite for leukotoxin HlgB/HlgC. However, the relationship between membrane PiP2 content and the regulation of SOCE (Korzeniowski et al., 2009;Walsh et al., 2009), could also be an alternative explanation for such inhibition.
The contribution of intracellular acidic stores as an early step of Ca 2+ mobilization induced by a bacterial toxin is a novel finding. It is worth mentioning that signalling molecules linking plasma membrane receptors to the release of Ca 2+ from internal stores are now starting to be identified (Cosker et al., 2010). The more potent of these molecules is NAADP, which is synthesized by the ADP-ribosyl cyclase CD38 (Malavasi et al., 2008) and activates the two-pore Ca 2+ channels from lysosomes (Zhu et al., 2009). Our various approaches in shutting down CD38 activity by mimicking published protocols (Zocchi et al., 1999;Ferrero et al., 2004;Hara-Yokoyama et al., 2008), successfully reduced or suppressed the leukotoxin effect, strengthening the probable relationship between leukotoxin HlgC/HlgB binding and Ca 2+ release from acidic stores. Summarizing the pharmacological manipulations that alter the granular neurones response to HlgC/HlgB leukotoxin, Fig. 9A illustrates a multistep process evoked by the endolysosomal Ca 2+ signalling (Morgan et al., 2011). The cell membrane molecule (or molecules) that bind the toxin needs to be identified; here is represented by a 'schematic receptor'. The step 'binding' (1) must be connected to the activation of the CD38 ADP-ribosyl cyclase, since the toxin activity is significantly inhibited by b-NAD acting on CD38 or the action of an anti-CD38specific antibody.
Step 2 comprises the NAADP mediated activation of Ca 2+ release from the endolysosomal compartment, as suggested by the total inhibition caused by Bafilomycin A or GPN. The local release of Ca 2+ triggers further Ca 2+ release from the thapsigargin-sensitive reticular store (step 3). The final step (4) involves the activation of CRAC channels induced by the Ca 2+ reticular stores depletion.
It would be of interest to determine whether other bacterial toxins mobilizing intracellular Ca 2+ initiate a related mechanism, regardless of other effects they may induce. Similarly to S. aureus bi-component leukotoxins that evoke intracellular Ca 2+ rise in neurones, even though they are known to discriminate between target cells (Menestrina et al., 2003). Irrespective of their similarities, the g-haemolysins HlgA-HlgB and HlgC-HlgB share their F-component, the S-components HlgC and LukS-PV can compete for a same binding site (Prévost et al., 1995;Meyer et al., 2009), S. aureus leukotoxins share an initial intracellular Ca 2+ mobilization which may induce a variable final outcome depending on the cell type or on the specific toxin. A question to focus on will be how bi-component bacterial toxins initiate cell signalling. It is tempting to speculate about some multi-molecular system, which should integrate the characterized cellular responses that are independent of the pore forming ability, such as cytotoxicity (Genestier et al., 2005;Zivkovic et al., 2011), enhanced bactericidal capacity (Graves et al., 2012) and Ca 2+ mobilization from acidic stores. A comparative study of the responses of cultured neurones, alveolar macrophage and neutrophils to the S. aureus bi-component leukotoxins would be of valuable interest.

Ethic statement
All experiments have been conducted using protocols designed according to the European and French guidelines on animal experimentation and approved by the direction of the Bas-Rhin veterinary office, Strasbourg, France; authorization number 67-312 to E.J.

Animals
Sprague-Dawley rats were obtained from the Institut des Neurosciences Cellulaires et Intégratives (INCI) animal facilities . All procedures regarding housing, feeding and sacrifice of the animals were in accordance with approved European guidelines (No. 86/609/CEE) on animal care and experimentation.

Glucose-6-phosphate dehydrogenase
Quantification of cell death was performed using the fluorescence-based micro-plate assay (Batchelor and Zhou, 2004) which measures glucose-6-phosphate dehydrogenase (G6PD) activity released into the media. The same supernatant collected for glutamate determination was also used for G6PDH activity. Reactions were conducted in a 100 ml final volume (96well plates). Experimental samples were mixed 1:1 (v/v) with reagent mix containing 15 mM resazurin, 2 mM Glucose-6-Phosphate, 0.5 mM NADP and 0.5 U ml -1 diaphorase in 100 mM Tris-HCl buffer, pH 7.5. The reaction was carried out at 37°C for 30 min and the increase in fluorescence was measured as described for the glutamate assay.

Immunocytochemistry
The subcellular localization of leukotoxin in granular neurones was detecting using the HlgB subunit tagged with Alexa Fluor 488 C5 maleimide (Life Technologies). Cells were incubated at room temperature for 1-10 min in the presence of 1 nM native or 5 nM tagged HlgC/HlgB. Neurones were then fixed in 4% (v/v) paraformaldehyde and 4% (w/v) sucrose in phosphate-buffered saline (PBS) for 15 min at room temperature and subsequently rinsed with PBS. Neurones were permeabilized and non-specific epitopes were blocked in PBS buffer containing 0.1% (v/v) Triton X-100, 1% (w/v) bovine serum albumin and 5% (v/v) normal goat serum (Chemicon-Millipore, Molsheim France) for 30 min prior to 2 h incubation at room temperature with antibodies diluted in the same buffer. The following primary antibodies were used: the anti-STIM1 (N-19) goat polyclonal antibody and A8 mouse monoclonal antibody; the anti-mouse CD38 (2Q1628) rat monoclonal antibody and M-19 goat polyclonal antibody (Santa Cruz Biotechnology). The plasma membrane was labelled by the cholera toxin B subunit conjugated to Alexa Fluor 594 (Life Technologies). After three washes in PBS, secondary antibodies tagged with Alexa 488 or Alexa 546 (Molecular Probes/Life Technologies, 1:1000) were incubated in PBS for 2 h at room temperature. After the immunostaining process, cell nuclei were stained using Hoechst 32258 (Sigma) at a final concentration of 40 nM. After washing in PBS, coverslips were stored at 4°C until observation. A Leica SP5-II confocal microscope was used.

Microfluorimetry
Intracellular [Ca 2+ ] recordings were performed on cells grown on glass coverslips and loaded with 5 mM Fura-2 AM -0.04% Pluronic 127 for 30 min in the dark at 37°C, washed twice in HBSS and transferred to an inverted epifluorescence microscope (Axiovert, Zeiss, Germany) equipped with a UPlanFL 40/0.75 objective. The cells were alternatively illuminated at 350 nm and 380 nm and image pairs of the 520 nm light emission were recorded every 2 s for 30 min. Cells were bathed in HBSS buffer containing 10 mM Hepes buffer (pH 7.2) or, if needed, continuously superfused with the same buffer at a rate of 1 ml min -1 . The toxins were applied directly into the bath. For DRG neurones, a solution of ATP (1 mM final concentration) preceded of 1 min the administration of the toxin; this ATP application never induced [Ca 2+ ]i changes. The ratio of fluorescence intensities (Ex350 nm/Ex380 nm) was calculated on a pixel basis for each image pair. The calcium concentration was then estimated by the formula (Grynkiewicz et al., 1985): where Kd is the dissociation constant of Fura-2 for Ca 2+ (924 nM, calculated in our cultures), b = (I380 max)/(I380 min) (2.08 Ϯ 0.03, n = 39). Values of Rmin and Rmax were determined by periodical calibrations (Grynkiewicz et al., 1985).

Statistics
Data are presented as box-plots with median, 10th, 25th, 75th and 90th percentiles and the outlying points; the mean line is also shown. Statistical significance was determined by Kruskal-Wallis one-way analysis of variance (SigmaPlot for Windows version 11.0). Differences were considered significant at P < 0.05. facility of the IFR37. Special thanks to Dr Nancy Grant (CNRS UPR3212, Strasbourg) for her valuable comments and discussions.
The experimental work was supported by grants from EA-4438 and Scientific Council from the Université de Strasbourg, and from UMR 3212. The funding agencies had no role in any of the study design, data collection and analysis, decision to publish or preparation of the manuscript. None of the results obtained are part of a patent or commercial product. M.Y.T. and B.-J.L. were supported by a doctoral fellowship from the Ministère de l'Enseignement Supérieur et de la Recherche and from the Fondation Fouassier respectively. E.J. and B.P. are permanent staff of the CNRS (governmental) and G.P. is a permanent faculty member of the Université de Strasbourg. The above does not alter the authors' adherence to all OnlineOpen policies on sharing data and materials.