Human neutrophils communicate remotely via calcium-dependent glutamate-induced glutamate release

Summary Neutrophils are white blood cells that are critical to acute inflammatory and adaptive immune responses. Their swarming-pattern behavior is controlled by multiple cellular cascades involving calcium-dependent release of various signaling molecules. Previous studies have reported that neutrophils express glutamate receptors and can release glutamate but evidence of direct neutrophil-neutrophil communication has been elusive. Here, we hold semi-suspended cultured human neutrophils in patch-clamp whole-cell mode to find that calcium mobilization induced by stimulating one neutrophil can trigger an N-methyl-D-aspartate (NMDA) receptor-driven membrane current and calcium signal in neighboring neutrophils. We employ an enzymatic-based imaging assay to image, in real time, glutamate release from neutrophils induced by glutamate released from their neighbors. These observations provide direct evidence for a positive-feedback inter-neutrophil communication that could contribute to mechanisms regulating communal neutrophil behavior.


INTRODUCTION
Neutrophils provide the first line of defense against pathogens during acute inflammation. [1][2][3] In humans, a significant decrease in neutrophil numbers could lead to severe immunodeficiency or death. These polymorphonuclear leukocytes mobilize to the site of inflammation and engage several cell-killing mechanisms to clear the infection. 4,5 Cellular mechanisms underpinning communication of neutrophils with other actors of a pathological immune response, such as platelets, T cells, or infectious agents, have been intensely studied. [6][7][8] Nonetheless important is the rapidly emerging knowledge of intercellular communication among neutrophils themselves. Cooperative, swarm-like migration patterns of neutrophils have been considered an essential process in their tissue response. 9,10 The underpinning molecular mechanisms involve the lipid LTB4 and integrins, the release of signaling molecules such as ATP, 11 and the action of connexins accompanied by cooperative calcium alarm signals. 12 Ca 2+ signaling has long been considered key to the physiological functions of neutrophils, 13 which are equipped with a variety of membrane receptors, including GPCRs, FcRs, and integrins, capable of mediating Ca 2+ messages. 14,15 Ca 2+ -dependent release of chemo-attractants enables self-sustained paracrine signaling, thus providing positive-feedback amplification that drives self-organized neutrophil ensemble behavior. 16,17 Generated in a small group of clustering neutrophils, the molecular signal thus triggers what could be a chain-reaction mechanism communicating with more distant cells attracting them for further swarm growth. 17 However, direct evidence for regenerative-type neutrophil-neutrophil communication at the cellular level has been scarce.
Previous work has shown that neutrophils express ionotropic glutamate receptors of NMDA type (NMDA receptors, NMDARs) and can secrete glutamate and the NMDAR co-agonist D-serine. 18,19 Because NMDAR activation can generate major Ca 2+ influx and thus engage release machinery in the host cell, we thought it was important to understand whether glutamate-induced glutamate release could contribute to the positive-feedback signal amplification among neutrophil ensembles. To explore this, we engaged single-cell neuroscience techniques that we previously established while exploring signal exchange among central neurons and astrocytes. 20 We used freshly isolated human neutrophils semi-suspended in a culture preparation that allows for some natural morphological plasticity of these cells (STAR Methods, Video S1). First, to probe directly the function of NMDARs expressed in neutrophils, 19 we held individual cells in whole-cell mode (current clamp; initially in 0-Mg 2+ solution). A 1-s pulse of NMDA (100 mM) and the NMDAR co-agonist glycine (1 mM) applied using a piezo-driven theta-glass rapid-solution-exchange system ( Figure 1A) 22 evoked inward current ( Figure 1B, top); it was blocked in the same cell either by extracellular Mg 2+ (2 mM), by the specific NMDAR antagonist 2-amino-5-phosphonovaleric acid (APV, 50 mM), or by the selective antagonist of GluN2B-containing NMDARs Co101244 (1 mM) applied consecutively every 10-20 s through a multi-channel system ( Figures 1B and 1C).
In separate experiments, we incubated neutrophils with the Ca 2+ indicator Fluo-4 and morphological tracer CellTracker Red (STAR Methods), to monitor their Ca 2+ level at each experimental epoch (after adjusting the focal position under two-photon excitation). The agonist stimulus led to a prominent Ca 2+ rise in control conditions ( Figure 1D), but was blocked when 2 mM Mg 2+ or Co101244 were present in the medium from the start ( Figures 1E and 1F). Toward the end of such tests, we confirmed the neutrophil Ca 2+ signaling capacity (to avoid false negatives) by applying the protein kinase C activator phorbol myristate acetate (PMA, 1 mM), which triggered a further Ca 2+ elevation in each such case ( Figures 1E and 1F).
These observations provide direct evidence for functional NMDARs in neutrophils, predominantly GluN2B subtype, revealing an important mechanism of Ca 2+ mobilization in these cells. We also used super-resolution microscopy (dSTORM) to further confirm significant expression of GluN2 subunits in human neutrophils ( Figure S1). However, neutrophil depolarization could also produce a robust Ca 2+ rise that was insensitive to APV ( Figure 1H), pointing to the additional, voltage-dependent and NMDAR-independent routes of Ca 2+ entry, such as Calcium release activated (CRAC) channels 15 or mechanoreceptor transient receptor potential melastatin (TRPM) channels ( 15,23,24 and references therein). : a neutrophil (held in whole-cell) is stimulated pharmacologically by applying different solutions through two channels of W-glass pipette (tip diameter $200 mm) mounted on a piezo-drive to enable the ultra-fast delivery (<1 ms resolution; solutions in W glass exchanged within 10 s using a rapid multi-channel perfusion system); see Videos S1 and S2 for live videos of neutrophils before and during patching.
(C) Summary of experiments shown in (B); mean G SEM (amplitude over the 300-500 ms pulse segment), normalized to control (sample size shown); ***p < 0.01. Inset, super-resolution dSTORM image of a neutrophil shown with chromatically separated GluN2B and elastase single-molecule labels as indicated; see Figure S1A for further detail and illustrations.
(D) Characteristic images of a neutrophil (gray DIC image, top raw) loaded with CellTracker Red (red channel, middle) and Second, to obtain further direct evidence for remote neutrophil-neutrophil interaction, we held two neutrophils positioned 5-15 mm apart, in whole-cell mode ( Figure 2A): these experiments were highly challenging as neutrophils tried to ''escape'' the patching pipette (Video S2) and stayed in whole-cell mode for several minutes only. Once in a dual-patch configuration, we applied a depolarizing stimulus to one cell to trigger its intracellular Ca 2+ rise (as Figure 1G). Remarkably, this evoked an APV-sensitive inward current in the other patched cell ( Figure 2B). The response occurred 70-140 ms after the stimulus ( Figure 2C, left). With glutamate diffusivity of $0.7 mm 2 /ms, 25 diffusion theory indeed suggests a 50-150 ms post-release lag, depending on the separating distance, before glutamate concentration reaches its peak ( Figure 2C, right). Intriguingly, for the glutamate concentration to peak at the minimum level sufficient for NMDAR activation (0.5-1 mM) 5-10 mm away, the neutrophil should generate a 5-10 mM glutamate source emanating from its surface (see STAR Methods).
To understand whether this signal exchange triggers intracellular Ca 2+ mobilization, we monitored intracellular Ca 2+ in pairs or groups of neutrophils. The depolarizing stimulus applied to one cell triggered a transient Ca 2+ rise in its neighbors ( Figure 2D; characteristic traces shown in Figure S2A and Video S3). This rise was blocked by the specific NMDAR antagonist 3-((+)-2-carboxypiperazin-4-yl)propyl-l-phosphonic acid (CPP) but not by the broad-range metabotropic receptor antagonist alpha-methyl-4-carboxyphenylglycine (MCPG), although in the latter case the cell Ca 2+ response was significantly scattered (

Glutamate-induced glutamate release from neutrophils
Finally, we sought to determine whether glutamate released from one neutrophil prompts glutamate release from its neighbors. We first used a glutamate-specific biosensor 26 ( Figure 2F, left) to show that a depolarizing stimulus at one cell produces a micromolar-range glutamate response at 5-15 mm from it (Figure 2F, right). 18,19 Next, to visualize directly the fate of glutamate during neutrophil signaling, we employed an established enzymatic assay for extracellular glutamate imaging 27,28 ( Figure 2G, top). While the sensitivity chart for the assay in steady-state (equilibrated) conditions of glutamate application has been reported, 29 presently we deal with transient point sources of glutamate, with its concentration dropping sharply away from the release site. We, therefore, confirmed the robust sensitivity of the assay by imaging its fluorescence time course in the vicinity of a patch-pipette tip following a brief pressure puff of glutamate (5 mM at the source; Figure 2G, bottom; Figures S2B-S2D).
We next monitored pairs or small groups of neutrophils using the enzymatic assay and found that even a brief stimulation of one neutrophil (light touch of a patch-pipette) induced pronounced glutamate rises, both at the stimulated and the neighboring cell ( Figure 2H, Video S6). The level of fluorescence response in simulated or neighboring neutrophils (200-300% DF/F 0 , Figure 2I) was comparable with that in the  Figure S2D). This may suggest that the endogenously released glutamate could indeed reach a local level of several micromolar, near the surface of both stimulated (n = 17) and remote cell (n = 76; p = 0.362; Figure 2I). In the case when swarming neutrophils are in relative proximity of one another, this level appears not far from the concentration range of $8 mM considered optimal for neutrophil chemotaxis and cytoskeleton polarization. 30 Intriguingly, blocking NMDARs suppressed glutamate-induced glutamate release incompletely, with some signaling remaining ( Figure 2I, Video S7), although the latter could be partly the effect of the background glutamate diffusing from the stimulated cell.

Glutamatergic neutrophil-neutrophil signaling
Cooperative neutrophil behavior, such as swarming, appears to rely on Ca 2+ -induced release of chemo-attractants, with the paracrine-autocrine actions providing positive-feedback signal amplification. 16,17,31 This involves release of various signaling molecules that target their receptors on the neutrophil surface. 11,12 Previous studies have indicated that neutrophils express NMDARs and can release glutamate. 18,19 We, therefore, hypothesized that glutamate-induced glutamate release could contribute to the self-propagating, communal molecular signaling among neutrophils.
Our experiments demonstrated direct signal exchange among human neutrophils, which enacts through Ca 2+ -dependent glutamate release from one or more cells, activating NMDAR-mediated Ca 2+ entry and glutamate release in the neighboring neutrophils. To image glutamate-induced glutamate release in a direct manner, we employed an enzymatic assay developed earlier 32 and validated in astroglial cultures. 28,29 The assay solution may partly buffer the freely diffusing glutamate, depending on its spatiotemporal dynamics in relation to the enzymatic reaction kinetics. However, the fact that we could still see remote, NMDAR-dependent glutamate signals in the assay suggests that our observations could actually underestimate such signals. Given the strong glutamate release potential of individual neutrophils, these cells must also have a powerful mechanism of glutamate uptake, as reported earlier, 33 particularly because an excess of extracellular glutamate could hamper neutrophil cell-killing abilities. 34 Although we detected no significant role of metabotropic glutamate receptors in the Ca 2+ signal exchange among neutrophils, these receptors could still contribute to their migratory behavior in vivo. 35,36 Intriguingly, recent data point to a metabotropic, Ca 2+ -entry independent action of NMDARs interacting directly with mGluRs in neutrophils. 37 Thus, we cannot rule out residual contribution of native mGluRs to Ca 2+ signals in response to glutamate: the wide scatter of MCPG-dependent data ( Figure 2E, Video S5) suggests that the mGluR blockade could decrease Ca 2+ response in some cells.
Similarly, evidence has been emerging for the regulatory roles of mechano-sensing TRP channels in neutrophils, 38 in particular their potential contribution to neutrophil chemotaxis in vivo. 39 While the role of TRP receptors is outside the scope of the present study, it is possible that both patch-clamp interference and mechanical stimulation of individual neutrophils involves TRP activation leading to an NMDAR-independent Ca 2+ mobilization, as reported here ( Figure 1H). Notably, our key tests were carried out in low Mg 2+ because NMDARs are subject to the Mg 2+ block unless the cell membrane is depolarized. 40 It is, therefore, a plausible assumption that such depolarization in vivo might be prompted by TRP activation. Alternatively, the membrane potential of neutrophils, which is maintained mainly by their Na + /K + pumps, 41 can undergo drastic depolarization during neutrophil activation, 42 thus exposing NMDARs to direct activation by glutamate. The latter suggests that our observations might be particularly relevant to neutrophil behavior under activation.

Indicators of regenerative signal propagation
Overall, our observations point to glutamate-induced glutamate release as a mechanism that could provide regenerative, self-maintained signal propagation through the neutrophil ensemble. At the same time, our focus on the glutamate-release NMDAR-dependent neutrophil signaling does not exclude, but rather complements, other signaling mechanisms associated with self-organization and pattern behavior of these cells in vivo. Swarming and dynamic clustering have been considered an essential process in their tissue response. 9,10 In this respect, recent observations of neutrophil behavior in vivo report a cooperative intracellular Ca 2+ mobilization, or ''calcium alarm'' signal upon contact of neutrophils with their target. 12

Limitations of the study
While the present results provide strong support for the latter hypothesis, our approach has important limitations. Firstly, having neutrophils semi-suspended in culture is important for patch-clamp recordings and reliable Ca 2+ and glutamate imaging, but it restricts free cell movement, thus depriving us of exploring pattern behaviors of the cell ensemble. We also noted that a small fraction of neutrophils in our preparations displayed spontaneous Ca 2+ discharges. In the majority of cells, the Ca 2+ ''wave'' was triggered by the stimulus, but for the small minority, we could not determine whether their activity was spontaneous or stimulus-triggered. That the spontaneous Ca 2+ activity in some cells did not trigger Ca 2+ waves, or that some cells did not respond to the stimulus, suggests, as the most parsimonious explanation, that some neutrophils in our experiments did not possess the mechanisms of Ca 2+ -dependent glutamate-induced glutamate release. Secondly, to ensure consistency of our experiments, which requires reproducible conditions of cell stimulation across varied protocols, we employed well-controlled electrical or light-touch mechanical stimuli of individual neutrophils. Clearly, it would not be possible to avoid ''touch'' when using patch-clamp, whereas using ''touch'' alone was less perturbing but more difficult to interpret. We could, therefore, only assume that such stimulation could represent, at least partly, real-life scenarios in which neutrophils face foreign bodies causing an inflammatory response.

STAR+METHODS
Detailed methods are provided in the online version of this paper and include the following:    Data and code availability d Data reported in this paper will be shared by the lead contact upon request.
d This manuscript contains no original programming code.
d Any additional information required to reanalyse the data reported in this paper is available from the lead upon request.

EXPERIMENTAL MODEL AND STUDY PARTICIPANT DETAILS Preparation of human neutrophils
Neutrophils were isolated from blood samples obtained by venepuncture from anonymized healthy volunteers, males and females, British adults who self-reported their race as ''White'', ages from 30 to 45, according to protocols approved by the UCL Research Ethics Committee (UCL Queen Square Institute of Neurology, London, UK), with the informed consent of participants.
For the isolation of neutrophils, we used a method of dextran sedimentation and differential centrifugation through a Ficoll-Hypaque density gradient as described in detail elsewhere. 47 Briefly, a sample of the whole blood was suspended in sodium citrate solution, and a suspension of neutrophils was obtained by sedimentation in 2% dextran (Sigma, UK) in 0.9% NaCl for 45-60 min at room temperature. The neutrophil-enriched upper layers of the suspension were collected and centrifuged (1150 rpm, 10 min at 4 C). Residual erythrocytes were removed by hypotonic lysis, and the obtained suspension was further centrifuged (1300 rpm, 6 min at 4 C); the pellet was further re-suspended in PBS and purified by gradient centrifugation over Ficoll-Hypaque (Sigma, cat. 1077; 1500 rpm, 30 min at 4 C). The resulting pellet containing neutrophils was finally re-suspended in HBSS (0-Mg 2+ /0-Ca 2+ ); cells were plated on coverslips coated with poly-DL-Lysine (1 mg/mL) and laminine (20 mg/mL) and were maintained until used (37 C, 5% CO 2 ). With this protocol, neutrophils remained lightly attached to the coverslip, forming a semi-suspension: this partly restricted their movement while allowing for patch-clamp and fluorescence imaging experiments in individual cells. At the same time, the cells were not flattened by strong adhesion and showed considerable morphological plasticity on the microscopic scale. We thus identified healthy cells by their normal morphology, including pseudopodia motility, stimulus evoked transient Ca 2+ responses, normal NMDAR-dependent activity, and evoked glutamate release. The cultures contained a significant proportion of apparently healthy, stimulation-responsive neutrophils for up to 24 h, but the present data were collected within the fist several hours post-isolation.

Patch-clamp electrophysiology
Visualized patch-clamp recordings from the neutrophils were performed using a Multipatch 700B amplifier controlled by pClamp 10 The previously established fast-application system 22 included a theta-glass application pipette with $200 mm tip diameter attached to the PL127.11 piezo actuator driven by the EÀ650.00 LVPZT amplifier (both from PI Electronics). Routinely, one pipette channel was filled with the bath solution and the other channel was filled with the bath solution containing 100 mM NMDA and 1 mM glycine, or alternatively with NMDA and glycine, plus NMDAR blockers such as APV or CPP (the latter is considered potent and therefore is more frequently used when the extent of glutamate glutamate transients is not known). The pressure in the application pipette was regulated by a two-channel PDES-02DX pneumatic micro ejector (npi electronic GmbH) using compressed nitrogen. To test the effects of various ligands, the application solutions in both theta-glass pipette channels could be exchanged within 10-12 s during the experiment using dedicated pressurized micro-circuits. NMDAR ligands were applied in 400 ms pulses 5 s apart.

Diffusion time course estimates
The glutamate concentration time course C(r,t) at distance r from the glutamate-releasing cell was estimated, as a first approximation, using the classical equation for a small source in an infinite volume: Cðr; tÞ = C 0 V 0 8ðpDtÞ 1=2 exp À r 2 4Dt where D = 0.7 mm 2 /ms is glutamate diffusivity, and C 0 stands for glutamate concentration at t = 0 within small source volume V 0 . For the sake of simplicity, in the calculations we assumed that that volume from which glutamate is released can be represented by a 0.5 mm layer around a 6 mm wide spherical neutrophil.
To reach the minimum glutamate concentration required to activate NMDARs (0.5-1 mM), this equation predicts C 0 in the range of $5 mM (or equivalently, $10 mM within a 0.25 mM layer around the neutrophil source).

Two-photon excitation (2PE) fluorescence imaging
For live-cell imaging, neutrophils were loaded with a morphological tracer CellTracker Red (5 mM, Invitrogen) and Ca 2+ -indicator Fluo-4/AM (5 mM, Invitrogen) in the presence of Pluronic F-127 (0.02%, Invitrogen) for 10 min at 30⁰C. After incubation with the dyes, cells were washed out for 10 min in a medium containing (mM): 119 NaCl, 2.5 KCl, 1.3 MgSO 4 , 2 CaCl 2 , 26 NaHCO 3 , 1.25 NaH 2 PO 4 , 12 glucose (95% O 2 and 5% CO 2 ; pH 7.4, 290-300 mOsm). Imaging was performed in a medium of the same composition containing either 0 or 2 mM MgSO 4 , as specified, at 30-33⁰C. Imaging was carried out using an Olympus FV-1000MPE system optically linked to a Ti:Sapphire MaiTai femtosecond pulse laser (SpectraPhysics Newport) at l 2P ex = 800 nm, with appropriate emission filters (Fluo-4: 515-560 nm band; Cell tracker red: 590-650 nm band), as detailed previously. 48,49 For the time-lapse recordings, z-stacks of fluorescent images (containing 5-10 cells within the field of view) were collected in a 1-min increment for 2 min before (baseline) and upon application of NMDA (100 mM) and glycine (50 mM) (bath application for the next $3 min). At the end of each experiment, the protein kinase C activator phorbol myristate acetate (PMA, 1 mM) was added for 2 min as a functional test. For the analysis, the Fluo-4 signal (G, green channel) was normalized to the Cell tracker fluorescence (R, red channel), and changes in Ca 2+ level were first normalized as DG/R after background subtraction and next presented as DF/F 0 change in Fluo-4 fluorescence. Only cells displaying a stable baseline and robust (>2-fold) DG/R increase in response to the PMA test, with no Fluo-4 saturation, were included in the statistics.