Proteoglycans involved in bidirectional communication between mast cells and hippocampal neurons

Background Mast cells (MCs) in the brain can respond to environmental cues and relay signals to neurons that may directly influence neuronal electrical activity, calcium signaling, and neurotransmission. MCs also express receptors for neurotransmitters and consequently can be activated by them. Here, we developed a coculture model of peritoneal MCs, incubated together with dissociated hippocampal neurons for the study of cellular mechanisms involved in the mast cell-neuron interactions. Methods Calcium imaging was used to simultaneously record changes in intracellular calcium [Ca2+]i in neurons and MCs. To provide insight into the contribution of MCs on neurotransmitter release in rat hippocampal neurons, we used analysis of FM dye release, evoked by a cocktail of mediators from MCs stimulated by heat. Results Bidirectional communication is set up between MCs and hippocampal neurons. Neuronal depolarization caused intracellular calcium [Ca2+]i oscillations in MCs that produced a quick response in neurons. Furthermore, activation of MCs with antigen or the secretagogue compound 48/80 also resulted in a neuronal [Ca2+]i response. Moreover, local application onto neurons of the MC mediator cocktail elicited Ca2+ transients and a synaptic release associated with FM dye destaining. Neuronal response was partially blocked by D-APV, a N-methyl-D-aspartate receptor (NMDAR) antagonist, and was inhibited when the cocktail was pre-digested with chondroitinase ABC, which induces enzymatic removal of proteoglycans of chondroitin sulfate (CS). Conclusions MC-hippocampal neuron interaction affects neuronal [Ca2+]i and exocytosis signaling through a NMDAR-dependent mechanism.

temperature and pressure sensitive, degranulating in response to noxious physical stimuli through a mechanism dependent on transient-receptor-potential channel 2 [11]. In addition to inducing the release of preformed granule constituents, MC activation leads to de novo synthesis of many bioactive compounds, including lipid mediators such as leukotrienes or prostaglandins or a variety of cytokines, chemokines, and growth factors [10,12]. Overall, preformed granular material is released in minutes, while de novo materials are secreted within hours. Notably, MC activation does not necessarily lead to degranulation. For example, exposure of MCs to lipopolysaccharide can cause the release of cytokines without observable degranulation [13].
The mast cell-nerve relationship has been discussed extensively [14][15][16]. MCs are often found in close proximity to nerves [17]. When stimulated, messenger molecules released from nerves (e.g., neuropeptides) can elicit MC degranulation, but also mast cell-derived mediators (e.g., histamine, serotonin) can modulate neurotransmission [18], creating a bidirectional positive feedback-loop that may result in neurogenic inflammation [19]. MCs and their mediators are important conveyors of the communication between the innate enteric immune system and the enteric nervous system, and there is strong evidence for nerve-MC interaction, particularly in gastrointestinal diseases such as inflammatory bowel syndrome [20,21].
The functions of MCs in the brain are unknown. However, clear evidence indicates brain MCs are important contributors to neuropsychiatric states such as anxiety [22] and are able to modulate neuroinflammation, which may affect diseases such as multiple sclerosis and Alzheimer's disease [2,23]. The interaction between MCs and CNS neurons has been little studied. We examined the possibility of a bidirectional communication between hippocampal neurons and MCs through the use of a coculture model, where rat peritoneal mast cells, representative of connective tissue mast cells [24], are incubated with dissociated hippocampal neurons. A similar model was developed for the study of the cellular mechanisms involved in the associations between MCs and nerves in vivo [17] and MCs and sympathetic neurons [25,26]. Hippocampal neurons were activated by depolarization and MCs were immunologically activated (by antigen) as well as by the compound 48/80 and LPS. After selective stimulation of MCs and neurons, we used [Ca 2+ ] i imaging to record subsequent cell activation. Besides, a cocktail of MC secretory products was applied directly onto neuronal somata causing [Ca 2+ ] i transients and exocytosis. This response was NMDAR-dependent. We show for the first time bidirectional communication between hippocampal neurons and MCs and a role of MCs in neurotransmission.

Preparation of neuronal cultures
Hippocampal neurons were cultured from brains of P0-P1 newborn rats isolated by surgical separation of the cranial bones by natural suture. CA-1 and CA-3 hippocampal regions from both sides were isolated under a light surgical microscope and removed from the brain. The two hippocampi were immersed in 1 ml of enzymatic solution and kept under continuous agitation (80 rpm) for 40 min and centrifuged at 200×g for 5 min. The enzymatic solution contained 1 ml of DMEM (PAA Laboratories), 0.2 mg of cysteine, 1 mM CaCl 2 , 0.5 mM EDTA and 5 units/ml papain (Worthington). The enzymatic solution was previously bubbled with 95% O 2 and 5% CO 2 for 20 min. After centrifugation, the supernatant was removed, and 1 ml of inactivating solution was added to the tissue to prevent further digestion. The inactivating solution contained 800 μl of DMEM (Gibco), 200 μl FBS (BioWest), 200 units of penicillin (Gibco), 200 μg of streptomycin (Gibco), 2.5 mg albumin and 2.5 mg trypsin inhibitor (Sigma). After this process, the tube was centrifuged again at 200×g for 5 min and mechanically dispersed afterward with a 1-ml pipette tip five times. The supernatant was discarded and replaced with neuronal media, containing: 1 ml Neurobasal (Gibco), 20 μl B27 (Gibco) (50×), 10 μl Glutamax (Gibco) (100×), 20 units penicillin and 20 μg streptomycin. Neurons were plated at a density of 150,000 cells/well (12-well plates) over a single feeding layer of astrocytes grown on 15 mm ϕ coverslips. For the astrocyte feeding layer, the coverslips were previously treated with 1 ml of an acidic solution containing 17 mM acetic acid, 0.8 mg collagen (BD Biosciences), and 1 mg poly-D-lysine (Sigma). For the treatment, the solution was gently spread over the coverslips with a cotton bud and let to dry overnight and washed with water three times. The feeding layer was made of astrocytes from rat and seeded on the cover slips at a density of 40,000 cells/dwell for 5 days. The astrocyte feeding layer was preserved by adding uridine solution that prevents cellular division. The uridine solution contained DMEM, 20.4 mM uridine (Sigma) and 8.1 mM 5-phluoro-2-deoxy-uridine (Sigma). Neurons were cultured between 14-17 DIV in an incubator at 37°C with an atmosphere of 95% O 2 and 5% CO 2 .

Mast cell isolation and preparation of cocultures
MCs were isolated from the peritoneal cavity of 3-to 5-month-old rats following a procedure described in detail elsewhere [27]. The peritoneal wash fluid contains a large number of cells, a small fraction of which are MCs. Using a Percoll (Sigma-Aldrich) gradient, fully differentiated connective tissue peritoneal MCs were separated from other cells [28].

Statistical analysis
Differences between groups were tested for statistical significance using Student's t test for unpaired data when the data passed a normality test (Lilliefors-corrected Kolmogorov-Smirnov) and a Mann-Whitney U test when it did not [29]. The data were regarded as statistically different when P < 0.05. All statistical analyses were performed using SPSS Statistics 22 software (IBM). All data were plotted as the mean ± SEM. The presented data represent experiments performed with cells from at least three different cultures.

Bidirectional communication between hippocampal neurons and MCs in coculture
When in a coculture of mixed hippocampal neurons and MCs, loaded with FURA-2 AM, we stimulated a neuron by direct depolarizing stimuli applying a pressure pulse of K + -enriched solution (70 mM KCl with iso-osmotic reduction of NaCl) (Fig. 1a), a high [Ca 2+ ] i transient was evoked upon the onset of the stimulus in neuronal cell body, as it was expected (Fig. 1b). The peak amplitude (Δratio of F360/F380) of Ca 2+ response was of 0.55 ± 0.06 a.u. After a short latency, nearly 50% of surrounding MCs (a number of 50 MCs, out of 101) experimented [Ca 2+ ] i oscillations of variable pattern (Fig. 1b) and low amplitude (ΔF/F, 0.078 ± 0.009) (Fig. 1d). No calcium signal was observed when the depolarizing pulse was applied on a pure MC culture. Interestingly, after MC responses, a new but short [Ca 2+ ] i transient was observed in 16% of tested neurons (5 out of 31). This neuronal signal only occurred in neurons cocultured with MCs but not in isolated hippocampal neurons, and it could be reproduced for a second time when a new depolarizing pulse of high K + was again applied, suggesting that MC mediators feedback on neuronal activity. The amplitude of the second neuronal [Ca 2+ ] i peak was 0.039 ± 0.007. While MCs responses occurred 33.1 ± 2.9 s after the onset of first neuronal transient induced by depolarization, late Ca 2+ transients evoked in neurons after MCs response showed a similar delay of 30.4 ± 8.8 s (Fig. 1e).

Activation of hippocampal neurons by mast cell degranulation
The selective stimulation of MCs in close proximity to hippocampal neurons by compound 48/80 and antigen caused an immediate change in appearance of most MCs that led to degranulation within minutes. When MCs were activated with C48/80 (5 s application), a rapid Ca 2+ elevation was elicited (ΔF/F, 0.21 ± 0.03) that persisted at mid-levels for minutes (Fig. 2a). The effect of MC degranulation by C48/80 on Ca 2+ signals of neighboring neurons occurred fairly rapidly (delay, 10.3 ± 2.9 s) (Fig. 2f) and consisted in an acute Ca 2+ transient with a peak amplitude of 0.24 ± 0.13 (Fig. 2e). The percentage of responding neurons was totaled 43%. MCs previously sensitized with monoclonal IgE against dinitrophenyl (DNP) and stimulated with the multivalent antigen DNP-conjugated human serum albumin (HSA) showed a mild degranulation response  (Fig. 2b) with a peak amplitude of 0.14 ± 0.01. Subsequently, 25% of hippocampal neurons responded after MC activation, since Ca 2+ waves of variable amplitude (ΔF/F, 0.32 ± 0.08) (Fig. 2e) and delay (mean, 46.8 ± 22.6 s) (Fig. 2f) were observed. The application of LPS for 1 min to MCs only evoked a short burst of oscillations in [Ca 2+ ] i of low amplitude (ΔF/F, 0.06 ± 0.007) and took a long time to appear (Fig. 2c). MCs did not show morphological changes or apparent degranulation, and [Ca 2+ ] i signals from adjacent neurons remained unaltered. The possibility of direct effects of DNP or C48/80 on neuronal [Ca 2+ ] i signal was tested by applying these agents to hippocampal cultures in the absence of MCs. Neither DNP nor C48/80 alone had an effect on neuronal [Ca 2+ ] i signals. However, C48/80 has been said to exert a direct and strong excitatory action on enteric neurons and visceral afferents [30]. Therefore, in preventing potential direct effects of these agents on neurons, or indirectly, via mixed glia, we prepared a cocktail of secretory mediators from MCs (MCM). The cocktail of MC mediators (MCM) consisted of an extract obtained from degranulation by heat (53°C, for 1 h) of 10 6 MCs per milliliter. Application of this MCM via a local pipette directly onto a hippocampal neuron cell body evoked a successful activation, measured by [Ca 2+ ] i transients, in 92% of cells (n = 121) (Fig. 3a). Hippocampal neurons from a pure culture without glia were also able to react to the MCM similarly to neurons from mixed cultures. As a control, when the cocktail of mediators was obtained from MCs without degranulation (subjected to 37°C instead of 53°C, for 1 h), it failed to induce neuronal activation (n = 10). Therefore, hereinafter, we used the MCM obtained by heat (53°C). In comparison with pure MCM, application of high K + -enriched solution evoked higher [Ca 2+ ] i transients with larger peaks (ΔF/F, 0.55 ± 0.06 vs 0.47 ± 0.03; p = 0.021) (Fig. 3e)

MC mediators elicit synaptic transmission in hippocampal neurons
Transmitter release is tightly regulated by the presynaptic Ca 2+ transient [31,32]. Because MC degranulation is capable of evoking evident Ca 2+ transients in hippocampal neurons, we examined the transmitter of release using fluorescent imaging. Activity-dependent unloading of fluorescent styryl dyes (FM dyes) from recycling synaptic vesicles provides a direct measure of synaptic function at the level of single synapses [33]. Synaptic vesicles were loaded with FM1-43 dye for 60-s bath perfusion using 70 mM [K + ] solution. Subsequently, cells were continuously perfused for 5-10 min with 3 mM [K + ] basal solution in order to wash out excess FM dye from neuronal membranes. The cells showed a substantial amount of internalized probe (Fig. 4a, c) as a consequence of a first exo-endocytic process. Synaptic FM dye release was induced by applying 70 mM [K + ] solution (Fig. 4b) or MCM (Fig. 4d) for 5 s through a microperfusion pipette positioned at a distance of approximately 100 μm from the respective neuron. Destaining of synaptic FM dye clusters, defined by a monoexponential loss of fluorescence, was in both cases observed. The amplitude and the time course of synaptic FM dye release have been shown to provide direct readouts of the vesicle pool size and the release probability, respectively, at the level of single synapses [34]. Thus, in order to analyze release amplitudes (ΔF) and time constants (τ), FM dye release traces of individual synapses were fitted by a mono-exponential function (Fig. 4e, f). MCM was able to induce an evident FM-unloading indicating MC mediators can induce synaptic vesicle release and retrieval in cultured neurons. However, depolarization with 70 mM [K + ] solution induced a significantly higher (17.6 ± 1.4 a.u.; p = 0.002; n = 6 neurons) synaptic release amplitude than the MCM condition (8.7 ± 1.7 a.u; n = 14 neurons) (Fig. 4g), suggesting the cocktail of MC mediators results in a weaker stimulus and recruits a smaller pool of vesicles than the depolarizing stimulus. In contrast, the rate of release (time constants,τ) was similar (high K + , 12.06 ± 0.42 s; MCM, 11.44 ± 1.07 s) (Fig. 4h) indicating MCM has the same release efficacy of the individual synapses as depolarization with 70 mM [K + ] solution.

MC modulation of hippocampal neuronal function
Next, we investigated which mediators were involved in the MC-neuron signaling. MCs contain an array of chemical mediators that can be released to the extracellular medium upon activation. We focused on the preformed products contained in the granules since the brief delay in neuronal reaction after MC activation (≈ 30 s) suggests a rapid release of mediators by exocytosis. To elucidate which mediators were involved in the promotion of neuronal activation, we first directly stimulated neurons with bioactive monoamines (histamine, serotonin) and ATP. All of them failed to evoke a [Ca 2+ ] i transient in hippocampal neurons. Neuronal [Ca 2+ ] i signal remained at baseline even for long applications (data not shown), suggesting these mediators by themselves are incapable of eliciting the rapid response observed by MCM. Second, we selectively blocked PAR2 receptors with the antagonist FSLLRY-NH2 to determine the involvement of mast cell-specific proteases. We incubated cells for at least 10 min with FSLLRY-NH2 (400 μM) before proceeding to apply MCM. However, the drug barely modified [Ca 2+ ] i signal evoked by MCM (ΔF, 106.19 ± 8.13%; area under curve, 75.87 ± 8.16%; n = 66) (Fig. 5b). Lastly, to resolve the implication of highly anionic serglycin proteoglycans (PGs) containing glycosaminoglycan side chains of either heparin or chondroitin sulfate (CS) type in neuronal activation, we incubated the MCM with heparinase or chondroitinase ABC for 1 h at 37°C to break down the PGs of heparin and CS, respectively. Heparinase did not change the peak amplitude and area under curve of [Ca 2+ ] i transients respect to untreated MCM (ΔF, 100.09 ± 7.7%; area under the curve, 85.6 ± 10.3%; n = 57) (Fig. 5c). Conversely, chondroitinase ABC reduced the amount of [Ca 2+ ] i entry into the cell (Fig. 5d). Although [Ca 2+ ] i amplitude did not decrease significantly (ΔF, 88.81 ± 4.47%), [Ca 2+ ] i transient was narrower than that observed by untreated MCM, resulting in a lower area under the curve (63.3 ± 8.89%; p = 0.006; n = 57) (Fig. 5h), suggesting PGs may be in part responsible of MC-neuron communication. Although, we combined both enzymes and increased incubation time of MCM to 24 h at 37°C to make sure the complete degradation of PGs, we did not observe a stronger effect on [Ca 2+ ] i signal (data not shown). On the other hand, we decided to look into the mechanism that allows [Ca 2+ ] i to rise in neurons following MCM application. Accordingly, we examined the participation of glutamate receptors by using the NMDAR antagonist, D-2-amino 5-phosphonovaleric acid (D-APV), and the AMPA and kainate receptor blocker, 6-cyano-7-nitroquinoxaline-2,3-dione (CNQX). Cell  (Fig. 5e). In sum, the results of pharmacological experiments indicate that MC granule mediators elicit calcium transients and exocytosis by activating NMDA receptors in hippocampal neurons and partly, at least, PGs of CS rather than heparin contribute to the mediation of these effects on neuronal physiology.

Discussion
The results presented here constitute the first report of bidirectional communication between MCs and hippocampal neurons at cellular level. We used [Ca 2+ ] i imaging of hippocampal neurons and peritoneal MCs in a coculture model to record their activation after a selective stimulation of both cells. MCs were activated when neurons were stimulated by depolarization through a pulse of 70 mM [K + ] solution (Fig. 1). MC activation always occurred after neuronal response with a latency of 33 s which largely can reflect the diffusion time from neurotransmitters to MC target. In the present study, around 50% of all MCs in the field of view surrounding the stimulated neuron reply by means of a period of [Ca 2+ ] i oscillations. The failure to activate all MCs may be due to non-viable MCs, because of a lack of the necessary growth factors in the coculture that are required for mast cell survival (SCF, IL·3) [35]. Moreover, distinct populations of secretory vesicles containing diverse neuropeptides and transmitters have been reported in different types of inhibitory hippocampal neurons [36], and it is probable that MCs cannot respond to all of them. Interestingly, after MC responses, neurons were again able to evoke small [Ca 2+ ] i transients, suggesting MCs are able to release some of their mediators that feedback on neuronal activity. The short latency measured between MC response and the late onset [Ca 2+ ] i transient in neurons (≈ 30 s) suggests that involved mediators are preformed bioactive molecules released by exocytosis, because new synthesized lipid mediators and cytokines are generally released in the hours following mast cell activation [24]. MCs did not show an apparent change in morphology, which is indicative of full degranulation, rather reversible fusion or "kiss-and-run": a mechanism of exocytosis involving the formation of a narrow fusion pore connecting transiently granule and plasma membranes that result in a slow and partial release of granule contents [37][38][39]. In MCs, the "kiss-and-run" mode would allow the secretion of small molecules contained in the granule, such as amines, but the complete release of large material such as PGs by full fusion would require stronger activation and degranulation [40]. Moreover, a mode of transient fusion referred to as cavicapture, in which fusion pore dilation allows the loss of cargo while the omega shape of the fusing granule is preserved, has also been described in MCs [27] and could also explain our observations. Therefore, a minor stimulation of brain MCs without degranulation may be sufficient to activate hippocampal neurons. Anyway, this argument is not in accordance with the observed degranulation of MCs stimulated by antigen (DNP) and C48/80, an apparent requirement to evoke neuronal activation (Fig. 2). LPS did not evoke MC degranulation nor neuronal response.
The cocktail of MC mediators obtained by heat for 1 h probably contains the whole arsenal of inflammatory mediators and biologically active substances. A pulse of this MCM elicited a consistent and reproducible [Ca 2+ ] i signaling in hippocampal neurons (Fig. 3). When diluted, the cocktail was able to evoke a neuronal signaling up to a 1:10 dilution, which contains the mediators extracted from around 100,000 MCs per milliliter. Brain MCs are not numerous [5,41]. However, their number and distribution can dramatically change in response to a number of environmental stimuli such as trauma and stress [42,43]. In addition, the number of infiltrated MCs increases in patients with Alzheimer's disease compared to healthy subjects [44]. Likely, both amounts of activated MCs and level of MC activation may determine the type and amount of released inflammatory mediators and influence brain state.
Ca 2+ influx mediates synaptic release, which is physiologically triggered by trains of action potentials [45]. Similar to bursts of action potentials, high [K + ] treatment results in sustained elevations of the presynaptic Ca 2+ concentration. Ca 2+ transients result in FM dye destaining which represents the balance of vesicle fusion and vesicle endocytosis [46,47]. We used quantification of FM dye release induced by the cocktail of MC mediators in cultured hippocampal neurons to determine their contribution to transmitter release. By analyzing the magnitude of release (ΔF) in individual synapses we demonstrate that the MCM yielded synaptic release associated with FM dye destaining, which can be discriminated from the depolarization by 70 mM KCl (Fig. 4). The loss of FM fluorescence induced by MCM stimulation was 50% lower than that observed by depolarization. This difference is expected because [Ca 2 + ] i transients evoked by MCM also showed lower peaks than those evoked by depolarization.
Neuronal regulation of MCs involves classical neurotransmitters such as acetylcholine, neuropeptides such as substance P, CGRP, vasoactive intestinal peptide (VIP), and neurotensin [48]. For instance, there has already been reported bidirectional signaling between MCs and submucous neurons in the human gut and that both VIP and CGRP are the neuropeptides involved in intestinal MC activation [49]. In the hippocampus, neuropeptide Y is present in a subpopulation of nonpyramidal cells [50] and it might be responsible for MC activation by hippocampal neurons in our coculture model. It will be necessary to evaluate this hypothesis in the future. On the other hand, pharmacological dissection of MC mediators to discriminate those able to activate hippocampal neurons is quite complex. Despite that we select only preformed mediators to evaluate the involvement in neuronal activation, there are too many alternatives because MCs have the unique ability to store and release certain preformed cytokines and growth factors, including tumor necrosis factor (TNF) [51] and vascular endothelial growth factor [52]. Actually, TNF can be rapidly released (first 10 min from MC granules) and subsequently secreted along with other pro-inflammatory cytokines with a new synthesis after several hours [53]. If release of involved mediators occurs by "kiss-and-run", PGs are unlikely to evoke neuronal activation because the high molecular weight of PGs [54] demands the complete fusion of granules (as explained above). However, solely pre-digestion with chondroitinase ABC produced a significant reduction in [Ca 2+ ] i transients evoked by the MCM (Fig. 5). Peritoneal MCs contain both heparin and galactosaminoglycan dermatan sulfate (chondroitin sulfate B) in their granules as demonstrated by analysis of the GAGs by using nuclear magnetic resonance spectroscopy [55]. It was reported that chondroitin sulfate is able to elicit inward currents, cell depolarization, and calcium transients by acting on AMPA and kainate receptors of hippocampal neurons [56,57]; therefore, it is not surprising we found that pretreated MCM with chondroitinase ABC was able to reduce [Ca 2+ ] i transients in hippocampal neurons. However, in our experiments, NMDA receptors as opposed to AMPA and kainate receptors seem to mediate neuronal activation by MCM. Without considering new synthesized mediators, granules contain a cocktail of preformed mediators that can be rapidly released. Although, one by one, neither histamine, serotonin, nor ATP were able to elicit a [Ca 2+ ] i transient in neurons, the combined action of substances from activated MCs may be different from those of single agents and complicated interactions may occur between MC mediators and the potential neuronal targets. Indeed, histamine by itself did not evoke an observable rise in [Ca 2+ ] i signal. However, it was reported that histamine potentiates NMDA-mediated synaptic transmission and excitotoxicity in hippocampal neurons [58,59]. When histamine is co-released simultaneously with CS, as it should occur during full exocytosis, the effects on synaptic transmission in the hippocampus must be more robust and even toxic than when histamine is released solely by the more subtle mechanism of "kiss-and-run". Thereby, the combined effects of MC mediators can have unsuspected consequences on neuronal characteristics. It may be particularly decisive for the inflammatory state of the cerebral tissue the contribution of MC cytokines. MCs can be activated by various molecules through the toll-like, interleukin-1, and neurotransmitter receptors and secrete pro-inflammatory cytokines. TNF and interleukin-1 family cytokines play a crucial role in the pathogenesis and management of the CNS inflammatory response [53]. Interestingly, it has been proposed the use of IL-37 (a cytokine member of the IL-1 family with immunosuppressive and anti-inflammatory properties) in neuroinflammation to help to improve the course of many neurological diseases [60,61].

Conclusions
Here, we demonstrated bidirectional communication between MCs and hippocampal neurons with important effects on Ca 2+ influx and neurotransmission. Increasingly, new evidence points towards MCs in neuroinflammation and brain disorders such as cerebral ischemia, multiple sclerosis, Alzheimer's disease, or depression [62]. As MC is a long-lived innate immune cell that responds to its microenvironment through a range of surface receptors and even to toxic molecules such as amyloid beta peptides [63], the accumulation of toxics in the brain may create a positive feedback loop and a sustained activation of MCs that can result in neuronal [Ca 2+ ] i overload and long-term neuronal damage.