Neutrophil Swarming in Damaged Tissue Is Orchestrated by Connexins and Cooperative Calcium Alarm Signals

Summary

Neutrophils are major inflammatory cells that rapidly infiltrate wounds to provide antimicrobial functions. Within the damaged tissue, neutrophil migration behavior often switches from exploratory patrolling to coordinated swarming, giving rise to dense clusters that further disrupt tissue architecture. This aggregation response is self-organized by neutrophil paracrine chemoattractant signaling (most notably of the inflammatory mediator leukotriene B4 [LTB4]). The coordination mechanism and possible evolutionary benefits of neutrophil swarms are elusive. Here, we show that neutrophil swarms require mutual reinforcement of damage signaling at the wound core. New biosensors and live imaging in zebrafish revealed that neutrophil chemoattractant synthesis is triggered by a sustained calcium flux upon contact with necrotic tissue that requires sensing of the damage signal ATP. This “calcium alarm” signal rapidly propagates in the nascent neutrophil cluster in a contact-dependent manner via connexin-43 (Cx43) hemichannels, which are mediators of active ATP release. This enhances chemoattractant biosynthesis in the growing cluster, which is instrumental for coordinated motion and swarming. Inhibition of neutrophil Cx43 compromises clearance of wound-colonizing P. aeruginosa bacteria and exacerbates infection-induced morbidity. Thus, cooperative production of alarm signals among pioneer clustering neutrophils fuels the growth of dense antimicrobial cell masses that effectively seal off breached tissue barriers from opportunistic pathogens.

Graphical Abstract

Highlights

• •Calcium alarm signals in neutrophil clusters locally promote attractant synthesis
• •Calcium alarm signals are dependent on ATP sensing and contact with necrotic tissue
• •Clustering neutrophils initiate and propagate calcium alarm signals via Cx43 channels
• •Neutrophil swarms and Cx43 restrict wound colonization by opportunistic bacteria

Poplimont et al. show that frontline neutrophils clustering at wound sites centrally coordinate damage sensing, calcium alarm signals, and the escalation of the inflammatory response. Intracluster coordination of signaling requires connexins and leads to dense swarms, which protect the breached tissue from opportunistic pathogens.

Introduction

Tissue damage triggers rapid recruitment of immune cells, with neutrophils as prime infiltrators [1, 2]. This migratory response marks the onset of inflammation, which is essential for protecting the breached tissue from infection while the slow process of tissue repair unfolds. Neutrophils are instrumental for killing bacterial pathogens through phagocytosis, release of proteolytic enzymes, and reactive radicals [1]. However, prolonged neutrophil residence can cause collateral tissue damage, perpetuate inflammation, and delay tissue repair and restoration of homeostasis [1]. Chronic inflammation forms the basis of numerous diseases and can also be co-opted by cancer cells to favor tumor growth and metastasis [3, 4]. Tuning neutrophil accumulation to desirable levels is thus an important biomedical target, yet our basic understanding of how this response naturally escalates under physiological conditions remains limited.

Interestingly, although the initial steps in neutrophil recruitment are driven by extrinsic cues, the escalation phase of the response is largely self-organized. Tissue injury results in local release of primary damage cues (damage-associated molecular patterns or DAMPs) from necrotic cells, including ATP or formyl peptides, which are normally not present in the extracellular environment [1, 2, 5]. To a certain extent, these primary signals may act directly as chemoattractants by signaling through corresponding G-protein-coupled receptors (GPCRs) [5]. Beyond this, DAMPs and other physiological stresses cause secondary production of chemoattractants by local tissue cells, including chemokines or arachidonic acid metabolites [1, 6, 7]. Altogether, this cocktail of attractants promotes exit of neutrophils from the blood (extravasation) and biased directional motion (chemotaxis) toward the site of injury within minutes. Thereafter, neutrophil behavior can switch from mere chemotaxis to highly coordinated and unidirectional motion that culminates in dense clusters at the wound core [8, 9, 10]. This so-called “swarming” or “aggregation” behavior is self-organized, as it relies on paracrine release of the lipid attractant leukotriene B4 (LTB4) by neutrophils [6, 8]. The decision to release chemoattractant is thus critical for the ultimate scale of the response. However, it remains unclear how neutrophil activation and chemoattractant synthesis might be coordinated in individual neutrophils.

Coordination of signaling has been found to play a role in reminiscent phenomena in unicellular organisms. Upon starvation, unicellular slime mold amoebae aggregate into a multicellular migratory slug capable of seeking nutrients [11]. This is driven by initial production of the chemoattractant cyclic AMP (cAMP) in a single amoeba, which triggers further cAMP release in nearby cells, resulting in traveling waves of attractant. Coordination of this response toward a single center point requires periodic and polarized emission of signal [11, 12]. By analogy, neutrophils may require specific signaling dynamics to trigger swarming [6, 9, 13]. However, how the attractant production might be triggered and coordinated across single neutrophils in situ is unknown. Recent evidence shows that macrophages can prevent swarming by cloaking the wound area, suggesting neutrophil access to the necrotic site is important [14]. Another interesting clue is that a critical threshold of initial clustering at the site of damage correlates with subsequent swarming [15]. However, directly relating these observations to neutrophil activation and chemoattractant synthesis has been hampered by the lack of tools to monitor the relevant signals in vivo.

Here, we take advantage of the genetic and imaging amenability of zebrafish, in which neutrophil swarming is conserved. We visualized the intracellular events leading to LTB4 synthesis in individual neutrophils. This revealed that activation of LTB4 biosynthesis preferentially occurs in clustering neutrophils at the wound core rather than individual migrating cells. This activation is associated with distinct calcium alarm signals that are triggered by contact with necrotic tissue and propagated among clustering neutrophils. Formation and intracluster propagation of these calcium alarm signals are dependent on connexin-43 (Cx43) hemichannels, which allow ATP release from live neutrophils, leading to autocrine and juxtacrine amplification of damage signaling. This communication coordinates and amplifies calcium fluxes in the cluster, locally promoting attractant production. Inhibition of connexin-mediated communication suppresses swarming and increases wound susceptibility to infection, suggesting that neutrophil swarming may have evolved as a beneficial host defense mechanism.

Results

Distinct Calcium Signals in Clustering Neutrophils

Neutrophil swarming is conserved in zebrafish, a model ideally suited for imaging and genetic manipulation [9, 16]. To establish the role of neutrophil LTB4 production in this model, we generated a transgenic zebrafish line, Tg(lyz:lta4h-EGFP), expressing leukotriene A4 hydrolase (LTA4H), an enzyme that catalyzes the conversion of LTA4 into LTB4 [17], the final step in LTB4 biosynthesis. We used a previously validated translation-blocking morpholino [18] to suppress lta4h expression, and this led to reduced neutrophil accumulation in wounds (Figures S1A–S1C). In contrast, lta4h knockdown did not affect neutrophil accumulation in wounds of Tg(lyz:lta4h-EGFP) larvae (Figure S1D). This confirmed that neutrophil-derived LTB4 drives neutrophil accumulation at wounds, as observed in mammalian systems.

We next characterized the dynamics of neutrophil swarming in two types of wound models. Because swarms are more likely to occur under high neutrophil density [8, 15, 19], we visualized the behavior of neutrophils after acute laser wound injury at a site rich in neutrophils, the caudal hematopoietic tissue (CHT), using two-photon ablation (Figures 1A and S1E; Video S1). Within 5 min, neutrophils began migrating to the wound in a highly directional and coordinated manner, forming clusters at the wound core by 20 min (Figure S1F; Video S1). To quantify this, we measured the radial speed of neutrophils over time, which reflects the level of coordination of migration [8]. When cells move in synchrony in the same direction, the amplitude of radial speed of the population is high. Accordingly, we detected a marked wave of synchronous directional motion, peaking at 15 min post-wounding in laser wounds (Figure S1G; Video S1). This time course is comparable with the evolution of neutrophil radial speed in mouse laser wounds [8]. We compared this to neutrophil responses in mechanical wounds, executed at the ventral fin nearby the CHT. In this model, neutrophil swarms could be imaged no earlier than 15 min post-wounding and showed lower magnitude of clustering and more variable peaks of radial speed (Figures S1E–S1H; Video S1). We prioritized the laser wound assay to capture the swarm initiation and exploit the faster and less variable cell dynamics.

Figure 1.

Calcium Dynamics in Neutrophils during Swarming

(A) Schematic of a 3-day-post-fertilization (dpf) zebrafish larva showing the area of two-photon laser wound damage and imaging.

(B) Construct expressing GCamp6F under the control of the lysozyme C promoter (lyz). Ca²⁺ binding to the calmodulin (Cam) domain of GCamp6F increases EGFP fluorescence.

(C) Time-lapse sequence of two-photon confocal image projections showing neutrophils (color-coded for GCamp6F intensity) migrating from the caudal hematopoietic tissue (CHT) toward a laser wound (LW) (dotted line) at the ventral fin-CHT boundary (VF/CHT) in a Tg(lyz:GCamp6F) larva. The calcium wave (CW) is indicated in the second panel and the neutrophil cluster with a white arrow. Scale bar represents 50 μm.

(D) Quantification of mean GCamp6F intensity over time in all neutrophils in the field of view in the first 2 min post-wound. Intensity values were normalized to the mean intensity of segmented neutrophils prior to wound. Each line represents mean intensity per experimental larva.

(E) Quantification of GCamp6F intensity over time in neutrophils clustering at the wound versus migrating neutrophils beyond the wound (dotted line in C). Intensity values were normalized as in (D). n = 8 larvae in 8 experiments. Dotted square denotes the time-bin corresponding to analysis in (D).

(F) Normalized GCamp6F intensity in relation to the surface area of segmented neutrophils. Individual dots represent single neutrophils or clustered neutrophils at the wound (red) or migrating cells beyond the wound (blue). The maximum surface area for single neutrophils is indicated with dotted line. Data are from 8 larvae in 8 experiments.

(G) Comparison of GCamp6F intensity in single (<600 μm²) versus clustering neutrophils (>600 μm²). n = 924 single cells and n = 347 clustered cells from 8 larvae in 8 experiments; Mann-Whitney test.

In all panels, error bars represent SEM. ^∗∗∗p < 0.0002. See also Figure S1 and Videos S1 and S2.

Video S1. Neutrophil Swarm Dynamics in Laser Wounds and Mechanical Fin Wounds, Related to Figure 1

Neutrophils in Tg(mpx:GFP)ⁱ¹¹⁴ larvae responding to a laser-induced focal tissue damage in the caudal hematopoietic tissue-ventral fin boundary or focal mechanical wound at the ventral fin. Time from the initiation of imaging is indicated in minutes. In the case of mechanical wounds, imaging started 15 min post-wounding. Maximum intensity projections from two-photon microscopy z stacks are shown. Scale bar = 50 μm. Frame interval is 30 s.

LTB4 production requires calcium-dependent translocation of biosynthetic enzymes to membrane compartments where lipid metabolism takes place [20]. Intracellular calcium dynamics have been observed in zebrafish epithelial cells [7] and neutrophils migrating in a solitary manner [21], but not in swarming neutrophils. To characterize this, we generated a transgenic line expressing a sensitive calcium indicator, GCamp6F [22], in neutrophils, hereafter referred to as Tg(lyz:GCamp6F) (Figure 1B). We visualized the behavior of Tg(lyz:GCamp6F) neutrophils after acute laser wound injury and observed three distinct signals on the basis of GCamp6F fluorescence intensity (Figure 1C; Video S2): first, a brief, tissue-wide calcium wave immediately after wounding that dissipated within 30 s (Figures 1C and 1D), an anticipated transient response of tissue to injury [7, 23]. Subsequently, neutrophils in the CHT began migrating toward the wound and showed dynamic local fluctuations in calcium intensity (Video S2). In contrast, upon arrival at the wound core, neutrophils underwent a whole-cell, sustained calcium flux concomitant with clustering, which was of markedly higher amplitude than the low-level fluctuations in migrating neutrophils (Figure 1E; Video S2). A similar pattern of calcium signals was observed in smaller scale wounds (Video S2). Strikingly, the calcium fluxes rapidly propagated across clustering neutrophils, giving rise to a cellular mass with relatively sustained calcium signaling (Video S2). Quantification showed that the mean calcium intensity in clustering cells was sustained at high levels throughout the first hour post-wounding (Figure 1E). Moreover, calcium intensity showed a positive association with cluster size, as solitary cells had lower calcium levels than clustering cells (Figures 1F and 1G). Altogether, this evidence revealed distinct types of calcium signals in migrating cells versus clustering neutrophils.

Video S2. Calcium Signaling and 5-LO Translocation Patterns in Clustering Neutrophils, Related to Figures 1 and 2

The first three videos show neutrophils in Tg(lyz:GCamp6F) larvae responding to a laser-induced focal tissue damage in the caudal hematopoietic tissue-ventral fin boundary. The first two videos show large and small-scale clustering examples. Neutrophils in these videos are color-coded for intensity (left), or grayscale at high (middle) and low brightness (right) to enable visualization of calcium dynamics in migrating and clustering cells respectively. Maximum intensity projections from two-photon microscopy z stacks are shown. Frame interval is 30 s. Scale bar = 25μm (first video) and 50μm. The third video shows series of examples of neutrophils in different Tg(lyz:GCamp6F) larvae showing propagation of the calcium signal from one cell to another. Arrow indicates neutrophils with low calcium levels encountering neutrophils with high calcium levels. Time is indicated in minutes. Image dimensions in μm in x,y: Stack 1 42x32; Stack 2 40x30; Stack 3 79x67; Stack 4 53x56 ; Stack 5 25x26 ; Stack 6 137x103. The fourth and fifth videos show neutrophils in Tg(lyz:GCamp6F)(left) crossed with Tg(lyz:5LO-tRFP) (right) responding to a laser wound damage or mechanical ventral fin wound. A 5LO-tRFP translocation event is highlighted in slow motion in each movie, occurring shortly after (laser wound example) or around the same time (mechanical wound example) as a calcium flux. Spinning-disc confocal microscopy was used after wounding. Imaging started 10-15 min post -wounding. Scale bar = 15 μm. Frame interval is 30 s (fourth video) or 1 min (fifth video).

Activation of LTB4 Biosynthesis Occurs Preferably in Calcium-Fluxing Neutrophils within Clusters

The discovery of distinct types of calcium signals in swarming neutrophils prompted us to investigate which of these are consequential on LTB4 biosynthesis. The rate-limiting step in LTB4 biosynthesis is the translocation of 5-lipoxygenase (5-LO or ALOX5) to the nuclear envelope membrane, where it converts arachidonic acid into LTA4 [20]. LTA4 can be further processed to different metabolites, but neutrophils are geared to produce LTB4 [9, 24]. Thus, 5-LO peri-nuclear translocation provides a microscopically tractable readout to identify neutrophils with active LTA4/LTB4 biosynthesis (Figure 2A). To link 5-LO translocation with calcium signals, we generated a zebrafish line expressing fluorescently tagged 5-LO in neutrophils Tg(lyz:tRFP-5LO) and crossed this with Tg(lyz:GCamp6F) fish (Figure 2B). The distribution of 5-LO was constitutively nuclear, as indicated by co-localization with nuclear DAPI staining (Figures S2A–S2D). To improve resolution for these subcellular dynamics, we used spinning-disk microscopy following acute laser wounding by two-photon microscopy. We detected 5-LO translocation events in neutrophils within clusters (Figures 2C and 2E; Video S2). Though many of these events were likely obscured by overlapping cells within the cluster, the events that could be discerned were limited to a median distance of 20 μm from the wound center (Figures 2C and 2E). These translocating cells were also characterized by a markedly higher calcium level compared with non-translocating cells (Figures 2C and 2E; Video S2). As further evidence, we performed mechanical wounding in the ventral fin, as the less-compact clusters in this model facilitated the detection of 5-LO dynamics. We discovered the same trend, in that peri-nuclear 5-LO translocations were detected preferably among the clustering, calcium-fluxing cells, with a median distance of 5 μm from the wound (Figures 2D and 2F; Video S2). These data suggested that the specific calcium fluxes observed in clustering cells are associated with activation of LTB4 synthesis and that chemoattractant production is spatially favored in this cell group.

Figure 2.

Activation of LTB4 Biosynthesis Is Favored in Clustering Neutrophils

(A) Schematic of LTB4 biosynthesis. cPLA2 (calcium-dependent phospholipase A2) and 5-LO are recruited to the nuclear membrane and produce arachidonic acid (AA) and LTA4, respectively. LTA4 is metabolized into LTB4 by LTA4 hydrolase.

(B) Constructs for transgenic expression of a fluorescent fusion of 5-LO with tRFP in neutrophils (below). Schematic of neutrophil with 5-LO nuclear translocation is shown.

(C and D) Spinning-disk confocal projections of neutrophils in 3-dpf double-transgenic Tg(lyz:GCamp6F)xTg(lyz:tRFP-5LO) zebrafish larvae after two-photon LW in the ventral fin-CHT boundary (C) or mechanical wound (MW) in the ventral fin (D). Blue dotted lines indicate the wound area occupied by clustering neutrophils. Zoomed images of examples of neutrophils with 5-LO translocation are shown. Time in relation to translocation is indicated in minutes in the first example. Examples are from three (LW) or two (MW) different larvae. Scale bars represent 50 μm and 5 μm, respectively.

(E and F) Quantification of mean distance from the wound center (x; left) and normalized GCamp6F fluorescence intensity (right) for 5-LO-translocating cells versus non-translocating cells in laser wounds (E) and mechanical fin wounds (F). GCamp6F fluorescence intensity was normalized to the most fluorescent cell in the corresponding frame.

(E) n = 41 cells (for translocating cells, each dot is a cell; for non-translocating cells, each dot represents the mean of all non-translocating cells in the same field of view; left) and n = 31 cells from 8 larvae in 5 different experiments (right).

(F) n = 17 (left) and n = 16 (right) cells from 5 larvae in 3 different experiments. Colored dots represent examples shown in individual images in (D).

Wilcoxon matched-pairs signed rank test. Error bar represents 95% confidence intervals of medians. ^∗∗∗p < 0.0002. See also Figure S2 and Video S2.

Neutrophil Calcium Fluxes Are Triggered upon Contact with Necrotic Tissue or Pioneer Fluxing Neutrophils

Based on these observations, we set out to investigate the mechanism driving the 5-LO-associated calcium fluxes in clustering neutrophils. One possibility was that the calcium fluxes resulted from passive entry, if the fluxing neutrophils were themselves in the process of dying and had lost membrane integrity. Alternatively, the calcium fluxes could have been due to active calcium entry. To explore these scenarios, we imaged neutrophil wound responses in Tg(lyz:GCamp6F) in the continuous presence of propidium iodide (PI) in the bath of the larva, which selectively stains nucleic acids in cells with impaired membrane integrity. The dye cannot penetrate the skin, but superficial wounding permitted transient interstitial access and local staining (Figures 3A and S2E; Video S3). Dying neutrophils were distinguished by loss of GCamp6F signal followed by uptake of PI stain, around 20–60 min post-wounding (Figures S2F and S2G; Video S3). Interestingly, the dying neutrophils would typically eject themselves outside the cluster rather than disintegrating within the cluster (Video S3). We found that the percentage of these dying/apoptotic neutrophils within the cluster was relatively low (Figure S2H) compared with the number of cells fluxing calcium. This suggested that the trigger of the 5-LO-associated calcium signals was unlikely to be limited to passive calcium entry through loss of membrane integrity. On the other hand, neutrophils underwent a calcium flux concomitant with abrupt deceleration of migration upon direct contact with necrotic tissue (Figures 3B and 3C; Video S3) or after contact with other clustering neutrophils at the wound core (Figure 3D; Video S2). This suggested that the calcium fluxes were associated with neutrophil sensing of necrotic cells and/or contact with pioneer activated neutrophils.

Figure 3.

5-LO-Associated Calcium Fluxes Are Triggered upon Contact with Necrotic Cells or Neutrophils with Ongoing Fluxes

(A) Schematic of two-photon laser wounding in the presence of propidium iodide (PI).

(B) Time-lapse, two-photon confocal projection images of a GCamp6F-expressing (white) neutrophil (indicated with an arrow) entering a contact with PI⁺ cells/tissue (red) in a Tg(lyz:GCamp6F) larva; time in relation to the first frame is indicated in seconds. Scale bar represents 10 μm.

(C) Quantification of speed (blue) and normalized GCamp6F (gray) in neutrophils before and after contact with PI⁺ tissue. GCamp6F intensity was normalized as in Figures 1D and 1E. Dotted red line indicates time of contact. Pooled cell data from n = 23 cells in 7 larvae and 4 experiments are shown.

(D) Examples of cell contacts transmitting calcium fluxes. Each case is represented by time-lapse images of a non-calcium-fluxing neutrophil (arrow) contacting a fluxing neutrophil. Time in minutes is indicated relative to cell-cell contact. Scale bar represents 10 μm. The quantification of neutrophil transmission of calcium fluxes is indicated in Figure 6.

Error bars represent SEM. See also Figure S2 and Video S3.

Video S3. Dynamics of Calcium in Clustering Neutrophils in Relation to Cell Death, Related to Figure 3

The first movie shows neutrophils in Tg(lyz:GCamp6F) larvae (white) incubated in PI (red) responding to a two-photon laser wound tissue damage. A neutrophil contact with PI⁺ necrotic cells is highlighted during the video. Maximum intensity projection of z stacks from two-photon microscopy is shown. Scale bar = 25 μm. Frame interval is 30 s. The second movie shows a series of examples of neutrophils in different Tg(lyz:GCamp6F) larvae, two of which incubated with PI (red) (subvideos 2 and 3 in the sequence), showing neutrophil death/apoptosis and ejection from the cluster. Ejected neutrophils are labeled with a white arrow and neutrophils taking up PI are indicated with white arrows. Time is indicated in minutes. Image dimensions in μm in x,y: Stack 1 53x51; Stack 2 59x59; Stack 3 49x47; Stack 4 46x51.

Extracellular Calcium Entry and ATP-Gated Calcium Channels Promote Calcium Alarm Signals

We then investigated which signaling pathways might underpin the 5-LO-associated calcium signals in neutrophils. Neutrophils can generate cytosolic calcium signals through activation of intracellular calcium stores (store-operated calcium entry [SOCE]) triggered by the inositol triphosphate (IP₃) receptor in the endoplasmic reticulum (ER) and/or plasma membrane channels, including calcium-release activated channels (CRACs), transient receptor potential (TRP) channels, and ligand-gated calcium channels (such as the ATP-gated P2X channels) [21, 25, 26, 27]. To assess the contribution of these pathways, we applied corresponding inhibitors in the bath of zebrafish larvae prior to laser wounding (Video S4). We used SKF96365, which blocks various plasma membrane calcium channels, including TRP, voltage-gated, ion-gated calcium channels, and CRACs [21]. We also used 2-APB, which blocks intracellular calcium release via the IP₃ receptor [28]. SKF96365 inhibited neutrophil intracluster calcium fluxes and the radial speed of migrating cells (Figures S3A–S3C), similar as an inhibitor of LTB4 signaling (U-75302; Figures S3A, S3E, and S3F; Video S4) [29]. Neutrophil motility levels were higher in the presence of SKF96365 and U-75302 (Figures S3D and S3G), indicating a specific defect in the overall directionality and coordination of movement rather than a defect in locomotion. By contrast, the 2-APB inhibitor did not significantly affect intracluster calcium fluxes (Figure S3E; Video S4) but affected motility, as indicated by a significant decrease of cell speed regardless of direction (Figures S3F and S3G). This suggested that plasma membrane channels are involved in the intracluster 5-LO-associated calcium fluxes.

Video S4. Neutrophil Swarm Dynamics in the Presence of LTB4 and Calcium Signaling Modulators, Related to Figure 4

The first four movies show neutrophils in Tg(lyz:GCamp6F) larvae responding to a laser wound damage in the presence of ethanol control, 2-APB, U-75302, SKF96365. Two-photon microscopy was used for wounding and imaging. Scale bar = 50 μm. Frame interval is 30 s over 60 min. The following videos (5-7 in the sequence) show neutrophils in Tg(lyz:GCamp6F)/Tg(lyz:5LO-tRFP) larvae responding to mechanical ventral fin wounding in the presence or absence of the calcium ionophore A23187, calcium chelator EGTA and P2X1 inhibitor NF279. Spinning-disc microscopy was used for imaging. In video 5, 50 μM A23187 was added 45 min after the start of imaging (15 min post-wounding). Scale bar = 25 μm. Frame interval is 30 s. In video 6, 0.5M EGTA was added 40 min after the start of imaging (15 minutes post wounding). Scale bar = 25 μm. Frame interval is 2 min. In video 7, 10 μM NF279 was added 1h prior to wounding and imaging started 15 min post-wounding. Scale bar = 25 μm. Frame interval is 30 s.

To establish whether extracellular calcium entry is sufficient to trigger 5-LO translocation and neutrophil arrest, we added a calcium ionophore (A23187) after neutrophils initiated migration to the wound. For these experiments, we utilized the mechanical wound model to facilitate quantification of 5-LO dynamics. Within minutes after ionophore addition, neutrophils pre-clustering at the wound experienced an increase in amplitude of calcium fluxes, and this was followed by a broader formation of calcium fluxes in migrating cells further from the wound (Figures 4A and 4B; Video S4). The calcium fluxes correlated with a generic increase of 5-LO translocation across the population (Figure 4C; Video S4). Interestingly, the initial enhancement of calcium fluxes in the cells pre-clustering at the wound coincided with triggering of migration by neutrophils further away, before these latter cells were triggered to experience a calcium flux and arrest themselves (Figures 4B and 4D; Video S4). This suggested that intracluster calcium fluxes may be sufficient to trigger recruitment of nearby neutrophils, consistent with their ability to activate chemoattractant biosynthesis.

Figure 4.

ATP-Gated Calcium Channels and Extracellular Calcium Entry Promote 5-LO-Capacitating Calcium Fluxes in Neutrophils In Vivo

(A) Spinning-disk confocal projections from neutrophils in Tg(lyz:GCamp6F)xTg(lyz:tRFP-5LO) larvae responding to MW, before and after addition of 50 μM calcium ionophore (A23187). Time post-wounding is indicated inside the images. Arrows indicate translocation events. Red dotted line indicates area occupied by clustering neutrophils. Scale bar represents 50 μm. Cartoon indicates area of wounding.

(B) Normalized GCamp6F intensity over time in clustering versus migrating cells (cells within or beyond denoted red line in A). The time in relation to calcium ionophore addition is shown. n = 768–2,140 cells per bin (migrating) and n = 116–461 cells per bin (clustering) from 8 larvae in 3 experiments.

(C) Percentage of 5-LO-translocating neutrophils out of all neutrophils visible in the field of view. Each line represents an individual larva; 8 larvae from 3 different experiments.

(D) Normalized radial speed over time for migrating cells. n = 743–1,851 cells per bin from 8 larvae in 3 experiments.

(E) Schematic of mechanical ventral fin wounding in the presence of NF279. Blue indicates vessels of the caudal vein plexus within the caudal hematopoietic site.

(F) Spinning-disk confocal projection images of neutrophils in Tg(lyz:GCamp6F)xTg(lyz:tRFP-5LO) larvae 120 min after MW in the presence (right) or absence (left) of 10 μM NF279. Red dotted line indicates area occupied by clustering neutrophils. Scale bar represents 25 μm.

(G) Mean normalized GCamp6F intensity larvae treated or not with NF279. GCamp6F intensity was normalized as in Figure 1D. n = 9 control larvae and n = 3 NF279-treated larvae from 3 and 2 experiments, respectively; Mann-Whitney test.

(H) Percentage of translocating neutrophils out of all neutrophils recruited into the fin over 2 h. n = 7 control and n = 3 NF279-treated larvae from 3 and 2 experiments, respectively; Mann-Whitney test.

Error bars represent SEM. ^∗p < 0.03, ^∗∗p < 0.002. See also Figures S3 and S4 and Videos S4 and S5.

To investigate whether extracellular calcium is required for intracluster calcium signals, we added a calcium chelator in the medium (EGTA) shortly after neutrophils started accumulating at the wound. This disrupted calcium dynamics both in clustering cells and in migrating cells (Figures S3H and S3I; Video S4). Speed of motion was compromised in the migrating cells upon depletion of calcium (Figure S3J; Video S4). This suggested that extracellular calcium is required for intracluster calcium signals but also for motility.

Next, we tested the contribution of plasma membrane channels more specific to damage sensing, such as the inhibitor for P2X1, an ATP-gated calcium channel that has been implicated in human neutrophil calcium fluxes [27]. We found that this inhibitor (NF279) markedly suppressed calcium fluxes and 5-LO peri-nuclear translocation in clustering neutrophils (Figures 4E–4H; Video S4). This suggested that the prominent calcium fluxes of clustering neutrophils at the wound core act as an alarm system that depends on exposure to the damage signal ATP (we hereafter refer to this as “calcium alarm signals”).

We then tested whether injection of individual chemoattractants, such as ATP, Cxcl8a, and LTB4, would be sufficient to trigger calcium alarm signals in vivo. We monitored neutrophil behavior and calcium dynamics 45 min after injection of these attractants in the otic cavity, an anatomical location devoid of neutrophils (Figures S4A–S4C; Video S5). None of these injections was sufficient to trigger sustained calcium alarm signals. We further investigated whether a cellular source of attractant would be effective in triggering such calcium fluxes. We monitored neutrophil behavior and calcium signals in the presence of Cxcl8a-mCherry-secreting transplanted cells, which we have previously shown to form extracellular chemokine gradients in vivo [30]. Neutrophils accumulated in the transplant but did not generate calcium alarm signals (Figures S4D and S4E; Video S5). This suggested that individual chemical signals, at least in the given configuration and doses, were insufficient to recapitulate the calcium fluxes seen at wounds.

Video S5. Neutrophil Calcium Dynamics upon Injection of Chemoattractants or Transplantation of Chemokine-Secreting Cells, Related to Figure 4

The first four videos show neutrophils in the head of 3 dpf Tg(lyz:GCamp6F) larvae responding to a local injection of PBS, 30 nM LTB4, 30 nM Cxcl8a and 200μM ATP. Spinning-disk confocal imaging started approximately 5 min after ear injection. Scale bar = 25 μm. Frame interval is 30 s. The last video shows neutrophils in the head of 3 dpf Tg(lyz:GCamp6F) (white) larvae responding to a transplant of Cxcl8a-mCherry-expressing HEK293T cells (red). Spinning disc microscopy was used one day post-transplantation. Maximum intensity projection of a z stack is shown. Scale bar = 25 μm. Frame interval is 20 s.

Neutrophil Cx43 Is Required for Coordinated Calcium Fluxes and Swarm Initiation

We had so far identified ATP sensing as a key damage cue that triggers calcium alarm signals in swarming neutrophils. However, this did not explain the highly efficient and coordinated spread of the calcium fluxes within the clusters. We hypothesized that clustering neutrophils may be mutually reinforcing ATP signaling. Neutrophils release ATP through connexin hemichannels, but whether this affects propagation of damage sensing has not been explored [31, 32]. To test this in vivo, we visualized neutrophil behavior in the presence of carbenoxolone (CBX), a drug that inhibits connexin channel activity [33]. This treatment profoundly inhibited neutrophil calcium alarm signals both in laser wounds (Figures 5A and 5B; Video S6) and mechanical fin wounds (Figures S4F and S4G; Video S6). Analysis of 5-LO events in the less-compact clusters of the latter model showed a reduced probability of 5-LO translocation after CBX treatment (Figures S4F and S4H). Further quantifications in our main laser wound model showed that CBX treatment led to exploratory single-cell motility, as indicated by high motility levels but low radial speed in comparison with untreated embryos (Figures 5C and S5A).

Figure 5.

Neutrophil Cx43 Is Required for Intracluster Calcium Fluxes and Swarm Initiation

(A) Time-lapse sequence of two-photon confocal image projections showing neutrophils in Tg(lyz:GCamp6F) larvae in the presence of PI, without treatment (ctr), with 50 μM CBX, or with morpholinos against cx43/cx43.4 (cx43 MO). Scale bars represent 50 μm and 10 μm for zoomed-out and zoomed-in images, respectively. Time after LW is shown in minutes. Arrows indicate neutrophils in contact with the wound.

(B) Normalized GCamp6F levels in control (n = 8), CBX-treated larvae (n = 5), and cx43 MO-treated (n = 7) from 8, 2, and 3 experiments, respectively. One-way ANOVA with Dunnett’s post test is shown. GCamp6F intensity was normalized as in Figure 1D. Data are from Tg(lyz:GCamp6F) and Tg(lyz:GCamp6F)xTg(lyz:tRFP-5LO) larvae evenly distributed across the groups.

(C) Neutrophil radial speed over time post-wounding for neutrophils in control, cx43 MO-treated, and CBX-treated larvae. Data are pooled from Tg(lyz:GCamp6F), Tg(lyz:GCamp6F)xTg(lyz:tRFP-5LO), and Tg(mpx:GFP)ⁱ¹¹⁴ zebrafish larvae. n = 1,201–1,719 cell steps per bin from 12 control, n = 1,268–1,535 cell steps per bin from 8 CBX-treated, and n = 1,308–1,554 cell steps per bin from 11 cx43 MO-injected larvae from 8, 3, and 4 experiments, respectively. Kruskal-Wallis with Dunn’s post test results indicating significance found between ctr and MO and between ctr and CBX (gray) or only between ctr and CBX (black) are shown.

(D) Time-lapse sequence of two-photon confocal image projections showing neutrophils in Tg(lyz:GCamp6F)xTg(lyz:cx43DN-T2A-mCherry) zebrafish larvae, positive (cx43 DN) or negative for the Cx43 DN-T2A-mCherry transgene (control siblings: Ctr sib). Annotations are as in (A).

(E) Normalized GCamp6F levels in neutrophils positive (cx43 DN) or negative for the Cx43DN-T2A-mCherry transgene (Ctr sib). n = 7 cx43 DN transgenics and n = 5 control siblings from 3 experiments; Mann-Whitney test.

(F) Neutrophil radial speed over time post-wounding for neutrophils positive (cx43 DN) or negative for the Cx43DN-T2A-mCherry transgene (Ctr sib). n = 1,330–1,602 cell steps per bin from 5 control siblings and n = 2,573–3,050 cell steps per bin from 7 Cx43 DN larvae from 3 experiments; Mann-Whitney test.

Error bars represent SEM. ^∗p < 0.03, ^∗∗p < 0.002, ^∗∗∗p < 0.0002. See also Figures S4–S6 and Video S6.

Video S6. Neutrophil Calcium and Swarm Dynamics with Cx43 Inhibition, Related to Figures 5 and 6

The first video shows neutrophils in Tg(lyz:GCamp6F) larvae (white) responding to a laser wound in the presence of 50 μM Carbenoxolone (CBX) and PI (red), imaged by two-photon microscopy. Scale bar = 25 μm. Frame interval is 30 s. The second video shows neutrophils in Tg(lyz:GCamp6F)/Tg(lyz:5LO-tRFP) larvae (white) responding to a mechanical wound in the presence of 50 μM Carbenoxolone (CBX), imaged by spinning-disk microscopy starting approximately 10 minutes after wounding. Scale bar = 50μm. Frame interval is 30 s. The third video shows neutrophils in Tg(lyz:GCamp6f) larvae (white) injected with a combination of cx43 MOs responding to a laser wound tissue damage in the presence of PI (red), imaged by two-photon microscopy. Scale bar = 50 μm. Frame interval is 30 s. The fourth video shows neutrophils in Tg(lyz:GCamp6F)xTg(lyz:cx43 dn-T2A-mCherry) zebrafish larvae, positive (cx43 DN; right) or negative for the Cx43 DN-T2A-mCherry transgene (control siblings; Ctr sib; left), responding to a laser wound and imaged by two-photon microscopy. Scale bar = 25 μm. Frame interval is 30 s. The fifth videos shows series of examples of neutrophils showing propagation of the calcium signal from one cell to another in different conditions: CBX treatment (first example), cx43 morpholino (second example) and cx43 DN (third and fourth example). Arrows indicate neutrophils with low calcium levels coming into contact with fluxing neutrophils. Time is indicated in minutes. Image dimensions in μm in x,y: example 1 61x62; example 2 75x75,; example 3 70x70; example 4 50x50.

To genetically corroborate these findings, we investigated connexin expression in purified zebrafish neutrophils and found two connexin genes to be expressed, cx43 and cx43.4 (Figure S5B). To assess cx43 and cx43.4 expression in neutrophils in situ, we performed whole-mount immunohistochemistry, using an antibody targeting both isoforms (Figures S5C and S5D). Cx43/Cx43.4 could be detected as vesicular puncta, usually in proximity to the membrane, in neutrophils of both wounded and unwounded larvae. Cx43/Cx43.4 was also detectable at the cell surface between contacting cells within clusters at the wound (Figures S5C and S5D). To inhibit cx43/cx43.4 expression, we tested combinatorial CRISPR-Cas9 knockout of cx43 and cx43.4. However, this led to zebrafish embryonic lethality before the onset of neutrophil development, consistent with the lethality of cx43-null mutations observed in mice [34]. We thus used knockdown with cx43 morpholinos that phenocopy hypomorphic mutations of cx43 [35]. The morpholino mixture for cx43/cx43.4 resulted in reduced retina size, an expected developmental phenotype [35] (Figures S5E and S5F). Consistent with the CBX results, cx43/cx43.4 knockdown reduced calcium fluxes in clustering neutrophils (Figures 5A and 5B; Video S6). cx43/cx43.4 knockdown also compromised swarming, as indicated by reduced radial speed but increased motion levels (Figures 5C and S5A; Video S6). To assess generic chemotaxis defects, we examined whether Cx43 inhibition compromises responsiveness to chemoattractants. We found that neutrophils in Cx43 morphants showed defective neutrophil recruitment to Cxcl8a injected in the otic cavity, but not to ATP and LTB4 (Figures S6A–S6D). This suggested specific cross-talks between Cx43/Cxcl8a signaling rather than generic defects in chemotaxis.

Given the broad expression of cx43, we next interrogated whether neutrophil cx43 is important for neutrophil swarming. To this end, we generated transgenic zebrafish whereby neutrophils express a dominant-negative version of cx43 (Tg(lyz:cx43DN-T2A-mCherry)), which inhibits Cx43 channel activity [36]. The behavior of neutrophils in these transgenics was similar as in cx43 morphants, in that they showed reduced whole-cell calcium fluxes and less-coordinated motility (Figures 5D–5F; Video S6). Assessment of neutrophil accumulation at fixed time points across a large pool of embryos showed that inhibition of neutrophil Cx43 suppressed neutrophil accumulation to a similar degree as global Cx43 inhibition (Figures S6E and S6F). This suggested that neutrophil Cx43 largely accounts for the overall defect in neutrophil accumulation at wounds. The accumulation defects were not due to suppressed motility, as neutrophil speed was higher in Cx43 DN mutants versus siblings (Figure S6G). Altogether, this evidence demonstrated an important role for neutrophil Cx43/Cx43.4 in coordinating intracluster calcium signaling and swarming.

Cx43 subunits can assemble into hemichannels that allow passage of ions and small molecules (including ATP) to/from the extracellular environment or into gap junctions that allow such conductivity across cells [37]. It remained unclear whether Cx43 mediates gap-junctional coupling or hemichannel-based ATP release and signaling [32, 38, 39] in clustering neutrophils. We did not observe PI uptake in live neutrophils, suggesting either absence of hemichannel activity or that the level of transport is below our detection limit (hemichannel opening could be transient and brief, unlike the permanent membrane integrity disruption in necrotic cells). We thus used functional tests to interrogate a link between Cx43 and ATP signaling in neutrophil swarming in vivo. Specifically, we found that Cx43 inhibition did not cause further reduction in neutrophil accumulation in NF279-treated larvae (Figures S6H and S6I). The absence of additive effects suggested that Cx43 and P2X1 may act in the same pathway, which would be consistent with hemichannel function of Cx43 in releasing ATP.

To better understand the role of Cx43, we quantified separately the pioneer calcium fluxes upon contact with necrotic tissue versus the subsequent transmission of these fluxes to other neutrophils. All types of Cx43 inhibition (CBX, cx43/cx43.4 knockdown, and expression of cx43 DN) reduced neutrophil calcium fluxes both upon contact with necrotic tissue and upon contacting other fluxing neutrophils (Figures 6A–6E; Video S6). This suggested that Cx43 likely plays a role in both autonomous and cooperative amplification of neutrophil damage signaling in the clusters. On the other hand, Cx43 inhibition did not affect the initial recruitment of pioneer neutrophils, as these arrived at the wound within the same time frame (Figure 6F).

Figure 6.

Neutrophil Cx43 Is Required for Autonomous and Cooperative Neutrophil Calcium Fluxes

(A) Time-lapse images showing calcium fluxes in neutrophils (arrow) contacting other neutrophils or necrotic cells in control untreated, CBX-treated, or cx43 MO-treated Tg(lyz:GCamp6F) larvae or in Tg(lyz:GCamp6F)xTg(lyz:cx43DN-T2A-mCherry) larvae. Time in minutes is indicated relative to the start of the cell-cell contact. Scale bar represents 15 μm.

(B) Cartoon illustrates contacts between neutrophils and necrotic PI⁺ cells or between fluxing and non-fluxing neutrophils resulting or not in calcium flux transmission.

(C and D) Percentage of neutrophil-neutrophil contacts (C) or neutrophil-PI⁺ cell contacts (D) resulting in transmission of calcium fluxes. Data are from Tg(lyz:GCamp6F)xTg(lyz:tRFP-5LO) (C and D) or Tg(lyz:GCamp6F) (D) larvae. Contacts in which none of the cells is initially fluxing are not included. n = 8 control, 4 CBX-treated larvae, and 7 Cx43 morphants from 8, 2, and 3 experiments, respectively (C). n = 5 control, 5 CBX-treated larvae, and 5 Cx43 morphants from 5, 2, and 2 experiments, respectively (D). One-way ANOVA, Tukey’s multiple comparisons test, is shown.

(E) Percentage of contacts resulting in transmission of calcium fluxes in Tg(lyz:GCamp6F)xTg(lyz:cx43DN-T2A-mCherry) zebrafish larvae, positive (cx43 DN) or negative for the Cx43DN-T2A-mCherry transgene (control siblings: Ctr sib). n = 7 cx43 DN and n = 5 control siblings from 3 experiments; Mann-Whitney test.

(F) Time point of arrival of first neutrophil at the wound in neutrophils positive (cx43 DN) or negative for the Cx43DN-T2A-mCherry transgene (control siblings: Ctr sib). n = 7 cx43 DN and n = 5 control siblings from 3 experiments; Mann Whitney test.

Error bars represent SEM. ^∗∗p < 0.002, ^∗∗∗p < 0.0002. See also Video S6.

Cx43 Promotes Wound Defense from Bacterial Invasion

In sterile injury, neutrophil swarms appear detrimental to tissue integrity, as they cause local tissue disruption [8, 9]. In certain parasitic infections, neutrophils promote pathogen spreading [40], and swarms could, in principle, facilitate this. The possible evolutionary benefit of neutrophil swarming is unclear, given these pathological implications. We hypothesized that dense neutrophil clusters, as opposed to diffuse patrolling, might provide a particularly effective antimicrobial barrier in breached tissue exposed to free-living opportunistic pathogens. To this end, we established a relevant wound infection model in zebrafish. Pseudomonas aeruginosa is an opportunistic bacterial pathogen that causes nosocomial wound infections [41, 42]. Wounds are nutrient-rich environments for these bacteria, which actively colonize these niches through chemotaxis [42, 43]. Acute wound infections by P. aeruginosa often spread rapidly, leading to sepsis and mortality within days or weeks [44]. We examined whether this is recapitulated in zebrafish by incubating larvae that had been mechanically wounded in a medium with PAO1 P. aeruginosa (Figure 7A). Non-wounded larvae were resistant to infection across a range of bacterial infection doses (Figure S7). In contrast, wounded larvae showed a dose-dependent decrease in survival (Figure S7), which correlated with an increase in total whole-body bacterial burden (Figures 7B and 7C). Using an intermediate infection dose, we found that cx43 morphant larvae showed significantly reduced survival and increased bacterial burden as early as 18 h after infection (Figures 7B and 7C). We obtained similar results by comparing infection-induced morbidity in lyz:cx43DN-T2A-mCherry larvae and control siblings (Figure 7D). These results demonstrated that Cx43 plays a role in restricting wound infections by pathogenic bacteria.

Figure 7.

Cx43 Is Required for Maximal Wound Defense from Bacterial Invasion

(A) Schematic of tail amputation and infection by PAO1 P. aeruginosa.

(B) Colony-forming units (CFUs) per larva in control wild-type (AB strain), non-injected larvae or cx43 MO-injected larvae. Time after wounding is in hours (h). n = 4 experiments, with 5 larvae per group; Mann-Whitney test.

(C) Survival over time in control wild-type, non-injected larvae or cx43 MO-injected larvae, wounded (W) or not (NW) in the presence of PAO1. n = 4 experiments, with 20 larvae per group; log rank (Mantel-Cox) test.

(D) Survival over time in Tg(lyz:GCamp6F)xTg(lyz:cx43DN-T2A-mCherry) zebrafish larvae, positive (cx43 DN) or negative for the Cx43DN-T2A-mCherry transgene (control siblings: Ctr sib), wounded (W) or not (NW) in the presence of PAO1. n = 4 experiments, with 20 larvae per group; log rank (Mantel-Cox) test.

(E) Schematic of imaging wound infection. Annotations are as in Figure 4E.

(F) Time-lapse sequence of two-photon confocal projections showing neutrophils in zebrafish larvae, positive (cx43 DN) or negative for the Cx43DN-T2A-mCherry transgene (control siblings; Ctr sib), in the presence of Syto62-labeled PAO1. Scale bars represent 50 μm and 10 μm. CW, CHT, and VF are as in Figure 1C. Arrows indicate neutrophils in contact with the wound. Dotted lines outline neutrophil clusters.

(G) Neutrophil cluster size over time post-wounding in cx43 DN or control larvae (Ctr) (includes negative siblings and single Tg(lyz:GCamp6F) transgenics). n = 5 cx43 DN and 7 control larvae from 6 and 5 experiments, respectively; Mann-Whitney test.

(H) Neutrophil radial speed over time post-wounding in cx43 DN or control larvae. n = 1,290–1,537 cell steps per bin from 7 control larvae and n = 1,527–2,079 cell-steps per bin from 8 cx43 DN larvae imaged in 6 and 5 experiments, respectively; Mann-Whitney test.

(I) GCamp6F levels normalized as in Figure 1D. n = 5 cx43DN transgenics and n = 7 control larvae from 6 and 5 experiments, respectively; Mann-Whitney test.

(J) Images of the wound (dotted outline) pseudocolored for fluorescence intensity of Syto62-labeled PAO1 bacteria. Time post-wounding is indicated in minutes. Scale bar represents 25 μm.

(K) Fluorescence intensity of bacteria at the wound relative to maximal initial intensity in this area. n = 5 Cx43 DN and n = 8 control larvae from 6 and 5 experiments, respectively; Mann-Whitney test.

Error bars represent SEM. ^∗p < 0.03, ^∗∗p < 0.002, ^∗∗∗p < 0.0002. See also Figure S7 and Video S7.

To link these findings with neutrophil swarm defects, we developed live imaging of wound infection by fluorescently labeled PAO1 P. aeruginosa (Figure 7E). Strikingly, wounding was followed by a sharp flow of bacteria toward the wound within 5 min post-wounding (Video S7). This was accompanied by neutrophil swarming with maximal clustering and cell coordination within 20 min (Figures 7F–7H), thus with comparable initiation kinetics as in sterile wounds. The neutrophil clusters were characterized by similar intracluster calcium signals as in sterile wounds (Figure 7I). To determine the function of clusters in bacterial fate, we analyzed the burden of bacteria in the area occupied by the neutrophil cluster (Video S7). We observed a rapid clearance of bacteria upon neutrophil clustering, as indicated by a decay in fluorescent signal (Figures 7J and 7K). Neutrophils expressing cx43 DN showed a delay in bacterial clearance (Figures 7J and 7K), consistent with defects in calcium alarm signals and neutrophil swarming in this time frame (Figures 7F–7I). Thus, the contribution of neutrophil Cx43 in wound immunity could at least in part be explained by its role in neutrophil clustering in this breached locus.

Video S7. Neutrophil Calcium and Swarm Dynamics during Wound Invasion by PAO1 Bacteria, Related to Figure 7

The first three videos shows neutrophils in Tg(lyz:GCamp6F)xTg(lyz:cx43 dn-T2A-mCherry) zebrafish larvae, positive (cx43 DN; S7B and C) or negative for the Cx43 DN-T2A-mCherry transgene (control siblings; Ctr sib; S7A), responding to a laser wounding in the presence of Syto62-labeled PAO1 bacteria (magenta), imaged by two-photon microscopy. Left images show merged green and magenta channels and right images show the magenta channel alone. Arrows indicate the wound core colonised by bacteria. Dotted lines delineate the larva and the incubation bath. Scale bar = 25 μm. Frame interval is 30 s. The last two videos show a 3D view of segmentation (magenta light) of PAO1 bacteria (magenta dark) at the neutrophil-occupied wound core (green), overlaid onto a sample of the first two movies. Scale bar = 50 μm.

Discussion

Neutrophil accumulation in inflamed tissue has pervasive implications in disease, and therapeutic strategies to fine-tune this process are desirable. A better understanding of how this migratory response physiologically escalates is important in this endeavor. Here, we reveal a cascade of signaling events underpinning amplification of neutrophil migration into prominent swarms. Our experiments reveal that Cx43 hemichannels drive coordinated calcium fluxes in a nascent neutrophil cluster, which promotes swarm growth by locally enhancing activation of chemoattractant biosynthesis. As Cx43 hemichannels mediate ATP release [27, 32], we propose that neutrophil Cx43 hemichannels amplify damage sensing in an autocrine and juxtacrine fashion at the wound focus to assemble a centralized, powerful chemoattractant gradient source. We show that formation of dense clusters through this mechanism is beneficial for early protection of the wound margin from opportunistic bacterial invasion.

Our study fills an important gap in our understanding of neutrophil swarming, as the basis of cell coordination has so far remained unclear. Our model provides a plausible explanation for why neutrophil access to the necrotic site is crucial for initiation of swarming [14] and why a primary neutrophil cluster precedes the onset of rapid aggregation [8, 15, 45]. The model also has implications in how an effective LTB4 gradient is self-generated by neutrophils during swarming [8]. The theoretical range of a gradient is dependent on the concentration of signal produced at the source and its diffusion and degradation rate [46]. The cooperative reinforcement of 5-LO-capacitating calcium fluxes combined with the rapid tissue diffusion of LTB4 [47] could play a part in the radius of the corresponding chemical gradient [8]. This does not exclude the contribution of LTB4-containing exosomes in the process, whose slower release and propagation could have additional effects in the response [48, 49]. Our data revealed a key role for neutrophil Cx43 in coordinating calcium alarm signals and swarm initiation. This mechanism could in part account for the reduced neutrophil accumulation in wounds of Cx43-deficient mice, a phenotype previously presumed to involve endothelial Cx43 [50].

Our wound infection data suggest an evolutionary benefit for neutrophil clustering. We found that the wound core is rapidly colonized by opportunistic bacteria within 5 min, which are largely cleared in this locus by neutrophil swarms within the following 20 min. This process was delayed when Cx43 and neutrophil swarms were inhibited. We speculate that suboptimal sealing of the wound by dense neutrophil clusters facilitates entry of bacteria into the deeper tissue. Such differences during the initial wound invasion could have knockon effects on subsequent host-pathogen interactions and, at least in part, account for the increased infection-induced morbidity in Cx43-inhibited larvae. The role of Cx43 and neutrophil swarming in wound defense could be relevant to other opportunistic bacteria capable of infecting wounds [51]. This does not exclude the possibility that other types of pathogens that propagate through neutrophils, such as Leishmania major [40], might exploit dense swarms for dissemination.

The coordination mechanism we describe is distinct from previous paradigms of collective cell migration. In a cohesive migrating group of cells, such as neural crest cells or the lateral line primordium, intercellular adhesion is critical for coordination of motion [52]. Cohesiveness further allows self-determination of directionality through asymmetric distribution of receptors across the moving cell mass [53, 54]. In the non-cohesive paradigm of slime mold aggregation, pulsatile release of attractant underpins coordinated gathering toward a single cell [11]. Here, we show that intercellular signal amplification within a seeding cluster powers the formation of a strong attractant source. This simple mechanistic principle appears to balance the benefit of rapid escalation with the risk of excess or misdirected congregation. The requirement for close cooperation in the primary cluster provides a level of stringency in the initiation of swarming. The preferential activation of attractant biosynthesis at the wound core provides spatial precision.

Our study points to interesting future lines of inquiry. One issue that remains unclear is why calcium alarm signals are spatially restricted in the wound-occupying cluster. P2X1 channel opening requires a threshold level of ATP, and such levels might be more likely encountered at the wound core if connexin hemichannels are selectively activated in this locus. Cx43 channel opening can be activated by various stimuli in vitro, such as LPS, LTB4, fMLP, and changes in extracellular ions [27, 32]. We reason that pioneer neutrophils contacting necrotic tissue initially experience death-specific signals (such as fMLP) that activate Cx43 opening. Subsequent calcium fluxes, 5-LO activation, and LTB4 release could then activate secondary Cx43 channel opening in a positive feedback loop [32]. This could account for the sustained duration of intracluster calcium fluxes and the inhibition of these fluxes by the LTB4 signaling blocker (U-75302). A positive feedback loop between LTB4 and Cx43 might also explain the suppression of intracluster calcium fluxes by the SOCE inhibitor SKF96365, as the latter blocks neutrophil activation by LTB4 [55]. Mechanosensation could be an intriguing cofactor in spatially restricting calcium alarm signals, as mechanical stimulation of the nucleus contributes to 5-LO activation [56]. A requirement for coincidence of multiple signals might explain why injection of individual chemical cues, including ATP, was insufficient to trigger prominent intracluster calcium fluxes in neutrophils. Another question raised by our study is whether centralized attractant production is sufficient to generate swarms and how many neutrophils might be required to form an effective gradient source. This would be interesting to address through in silico modeling and generation of optogenetic tools to manipulate such dynamics [57, 58]. Finally, a key point to elucidate will be the factors that terminate neutrophil swarming. Although the duration of 5-LO translocation was difficult to track in the dense clusters, it is noteworthy that translocation events and calcium fluxes were detectable beyond the duration of the migration wave. As monocyte/macrophage recruitment correlates with cessation of swarms in mice [8] and promotes neutrophil wound departure in zebrafish [59], it would be interesting to explore the role of these cells in catabolism of neutrophil-derived LTA4/LTB4.

Altogether, our study describes a novel mechanistic paradigm of collective cell behavior and identifies connexin channels as a key determinant of neutrophil swarming and wound immunity. This opens several avenues for investigation of this pathway in physiological and pathological conditions.

STAR★Methods

Key Resources Table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Chicken anti-GFP	Abcam	CAT#ab13970; RRID: AB_300798
Rabbit anti-tRFP	Evrogen	CAT#AB233; RRID: AB_2571743
Rabbit anti-Cx43	Sigma-Aldrich	CAT#C6219-1MG; RRID: AB_476857
Anti-chicken Alexa488	Invitrogen	CAT# A-11039; RRID: AB_142924
Rabbit anti-β-Tubulin	Abcam	CAT#ab209866
Rabbit IgG control	Merck	CAT#12-370; RRID: AB_145841
Goat anti-rabbit HRP antibody	Abcam	CAT#ab97080; RRID: AB_10679808
Anti-rabbit-Cy3	Jackson ImmunoResearch	CAT#711-165-152; RRID: AB_2307443
Bacterial and Virus Strains
PAO1 Pseudomonas aeruginosa	Martin Welch	N/A
Chemicals, Peptides, and Recombinant Proteins
Cxcl8a	ProteinTech	N/A
non-hydrolysable ATP-γS	Sigma-Aldrich	CAT#A1388-1MG
SKF 96365	Cayman Chemical	CAT#10009312
P. aeruginosa isolation agar	Scientific laboratory supplies LTD	CAT#257002
Nalidixic acid	Scientific laboratory supplies LTD	CAT#SR102E
DMEM	Invitrogen	CAT# 16219961
FBS	GIBCO ThermoFisher Scientific	CAT# 16140071
Penicillin/Streptomycin	Sigma-Aldrich	CAT#TMS-AB2-C
DAPI	Sigma-Aldrich	CAT#D8417-1MG
Lipofectamine 2000	invitrogen	CAT# 11668019
LTB4	Sigma-Aldrich	CAT#L0517-10UG
NF279	BIO-TECHNE LTD	CAT#1199
U-75302	Cayman Chemical	CAT#70705
MS-222	Sigma-Aldrich	CAT# E10521-50G
2-APB	Tocris Bioscience	CAT#1224
Methylene blue	Sigma-Aldrich	CAT# M9140-25G
1-phenyl-2-thiourea	Sigma-Aldrich	CAT#P7629-25G
EGTA	Sigma-Aldrich	CAT#E3889-25G
Carbenoxolone	Sigma-Aldrich	CAT#C4790-5G
Methanol-free formaldehyde	Thermofisher	CAT#28906
Sudan Black	Sigma-Aldrich	CAT#3080-1KT
PBS	Oxoid	CAT#BR0014G
Tween-20	Sigma-Aldrich	CAT#P-1379
Syto-62	Invitrogen	CAT#S11344
Low melting point agarose	Invitrogen	CAT#16520-100
A23187	Sigma-Aldrich	CAT# C7522-10MG
Critical Commercial Assays
KOD Hot start kit	Novagen, TOYOBO	CAT#71086
SuperScript III Reverse Transcriptase	Invitrogen	CAT#18080051
RNAeasy minikit	QIAgen	CAT#74104
Pierce ECL Plus Western Blotting Substrate	Invitrogen	CAT#32132
Sp6 mMessage mMachine kit	Ambion	CAT#AM1340
Experimental Models: Cell Lines
HEK293T cells	Felix Randow’s laboratory	N/A
Experimental Models: Organisms/Strains
Zebrafish Tg(lyz:GCamp6F)	This paper	N/A
Zebrafish Tg(lyz:cx43 dn-T2A-mCherry)	This paper	N/A
Zebrafish Tg(lyz:tRFP-5LO)	This paper	N/A
Zebrafish Tg(mpx:GFP)i114	[60]	N/A
Zebrafish Tg(lyz:lta4h-eGFP)	This paper	N/A
Zebrafish AB strain	PDN fish facility	N/A
Oligonucleotides
Primer cx43 Forward: 5¢-GCTCTCCA CTCTTTACTTCTTTCCAG-3¢	This paper	N/A
Primer cx43 Reverse: 5¢-GTATTGCACTTGAAAGCTGACTGC-3¢	This paper	N/A
Primer cx43.4 Forward: 5¢-GAGTCGTCATCGCGAGACATTGA-3¢	This paper	N/A
Primer cx43.4 Reverse: 5¢-GTCTATGAGTCTCAATCAAGCATGGATCC-3¢	This paper	N/A
Primer lta4h Forward: 5¢-TCTGAGAAGGAATATGTGGATGAA-3¢	[18]	N/A
Primer lta4h Reverse: 5¢-CAGCAAGAGATCTGTCTCCA-3¢	[18]	N/A
cx43 translation morpholino: 5¢-GTTCTAGCTGGAAAGAAGTAAAGAG-3¢	Gene Tools	ZFIN: ZDB-MRPHLNO-080818-1
cx43.4 Splice Morpholino: 5¢-ACTTCTCCATCTCCGTTATATTTTG-3¢	Gene Tools	N/A
Control Morpholino 5¢-CCTCTTACCTCAGTTACAATTTATA-3¢	Gene Tools	https://www.gene-tools.com/content/negative-control-morpholino-oligos
lta4h splice morpholino: 5¢-CAGTCTGATCAAGAGAAAGACTCGA-3¢	Gene Tools	ZFIN: DB-MRPHLNO-171122-3
Recombinant DNA
GCamp6F	[22]	N/A
lta4h-EGFP	Gene synthesized (Genewiz)	lta4h cDNA: ENSDART00000028171.7
tRFP-5lo	cDNA library from whole adult zebrafish	5lo/alox5a cDNA: ENSDART00000079884.6
cx43dn_T2a_mCherry	(Gene synthesized (Genewiz)	N/A
Cxcl8-mCherry	[30]	N/A
pCS2-TP	[61]	N/A
Software and Algorithms
MATLAB R2018b	https://www.mathworks.com/	N/A
Imaris v8.2	https://imaris.oxinst.com/	N/A
Fiji	https://fiji.sc/	N/A
Other
Bolt 10% Bis-Tris Plus Gel	Invitrogen	CAT#NW00100BOX
1mm glass beads	BioSpec	CAT#11079110
iBlot2 transfer stack	Life Technologies	CAT#IB23001

Resource Availability

Lead Contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, Milka Sarris (ms543@cam.ac.uk).

Materials Availability

Plasmids and zebrafish lines generated in this study are available upon request to the lead contact.

Data and Code Availability

The custom MATLAB codes are available at GitHub [https://github.com/LeukocyteMotionAndDynamics/NeutrophilSwarming] with sample datasets.

Experimental Model and Subject Details

Zebrafish were maintained in accordance with UK Home Office regulations, UK Animals (Scientific Procedures) Act 1986. Adult zebrafish were maintained under project license 70/8255, which was reviewed by the University Biomedical Services Committee. Animals were maintained according to ARRIVE guidelines. Zebrafish were bred and maintained under standard conditions at 28.5 ± 0.5°C on a 14h light: 10h dark cycle. Embryos were collected from natural spawnings at 4-5 hours post-fertilization (hpf) and thereafter kept in a temperature controlled incubator at 28°C. Embryos were grown at 28°C in E3 medium, bleached as described in the Zebrafish Book [62] and then kept in E3 medium supplemented with 0.3 μg/ml of methylene blue and 0.003% 1-phenyl-2-thiourea (Sigma-Aldrich) to prevent melanin synthesis. For live imaging of neutrophils expressing fluorescent markers, methylene blue was omitted from E3 medium to minimize tissue autofluorescence. All embryos were used between 2.5-3.5 dpf, thus before the onset of independent feeding. For live imaging or fixation, larvae were anesthetized in E3 containing 160-200mg/L MS-222 (Sigma). Where indicated, larvae were treated with 50 μM Calcium ionophore A23187 (Sigma), 10 μM NF279 (BIO-TECHNE LTD), 50 μM Carbenoxolone (Sigma), 3μM U-75302 (Cayman Chemical), 20μM SKF 96365 (Cayman Chemical), 25μM 2-APB (Tocris Bioscience), 0.5M EGTA (Sigma) or 1:1000 ethanol in E3.

Method Details

DNA constructs and transgenic zebrafish lines

Transgenic Tg(mpx:GFP)ⁱ¹¹⁴ zebrafish originated from the lab of S. Renshaw [60]. The new transgenic lines were made using a backbone DNA construct carrying a Lysozyme C promoter (lyz), for neutrophil-specific expression, minimal Tol2 elements, for efficient integration, and a SV40 polyadenylation sequence [63]. The references for the sequences cloned in this backbone vector are:

• GCamp6F: cDNA originally described by Chen et al., 2013 [22]

• lta4h-EGFP: lta4h cDNA (Ensembl: ENSDART00000028171.7) synthesized by Genewiz

• tRFP-5lo: 5lo/alox5a cDNA (Ensembl: ENSDART00000079884.6) amplified from a cDNA library from whole adult zebrafish

• cx43dn_T2a_mCherry: cDNA for cx43dn described by Omayada et al., 2002 [36] was synthesized by Genewiz

The sequence of zebrafish alox5/5-LO was chosen over 4 alox genes on the basis of similarity with human 5-LO [64]. For transgenesis, 0.5nL of solution containing 25ng/μL DNA plasmid and 35 ng/μL were injected into the cytoplasm of one-cell stage embryos. Transposase mRNA was synthesized from pCS2-TP [61] by in vitro transcription (SP6 message machine, Ambion). Injected embryos were stored at 28°C until 5dpf and thereafter were raised in the fish nursery according to standard rearing protocols. At 3 months old, F0 fish were outcrossed to a wild-type (TL) line in order to screen for germline transgenesis.

Morpholino injections

Morpholinos were ordered on GENE TOOLS LTD and their names, sequences, types and origins are indicated below. All morpholinos were injected in one cell stage eggs in a morpholino injection solution (120mM KCl, 20mM HEPES, 0.1% phenol red). We used morpholino against the two isoforms found in neutrophils: cx43 (also called cx43.3) [35] and against cx43.4. 1nL of 0.2mM of each morpholino was injected (0.4mM total). As injection control we used 0.4mM of Negative Vivo-Morpholino control oligo. For lta4h knockdown 3nL of 0.5mM MO LTA4H were injected [18]. List of morpholinos injected:

• cx43 (Translation-blocking): 5′-GTTCTAGCTGGAAAGAAGTAAAGAG-3′

• cx43.4 (Splice-blocking): 5′-ACTTCTCCATCTCCGTTATATTTTG-3′

• Standard negative control oligo: 5′-CCTCTTACCTCAGTTACAATTTATA-3′

• lta4h (Splice-blocking): 5′-CAGTCTGATCAAGAGAAAGACTCGA-3′

Western blotting

For western blotting, 10 larvae (3dpf) of each genotype were collected. Larvae were then lysed in 100μL of OCG buffer (0.3M NaCl, 2.5μM EDTA pH8, 0.9M Tris HCl pH7.5, protease inhibitors, phosphatase inhibitor) with 1mm glass beads (BioSpec) for 3x20sec in the sonicator Bioruptor (diagenode). Qubit protein assay kit (Invitrogen) was used to obtain protein concentration. Proteins (25μg) were resolved on a Bolt 10% Bis-Tris Plus Gel (Invitrogen), blotted onto nitrocellulose membrane using iBlot2 transfer stacks (Life Technologies) according to manufacturer’s protocol. Proteins were probed with rabbit anti-human Cx43 (1:2000) (Sigma-Aldrich) and rabbit anti-β-Tubulin antibodies (1:2000) (Abcam) after saturation in PBT (PBS, 0.1% Tween-20) containing 5% of milk. Proteins were then revealed using an enhanced chemiluminescence detection system (Pierce ECL Plus Western Blotting Substrate, Invitrogen) with goat anti-rabbit HRP antibody (1:2000) (Abcam).

Two-photon laser wound and live imaging

For mechanical ventral fin wounds, larvae were mounted immediately after wound onto a glass-bottom plate in 1% low melting agarose (Invitrogen) or a custom-built coverslip chamber (for when using an upright scope). Agarose-embedded embryos were covered with 2 mL E3 medium (supplemented with MS-222) and imaged either on i) an inverted PerkinElmer UltraVIEW ERS, Olympus IX81 spinning disk confocal microscope with a 30x/1.05 NA silicon (Olympus) or 40x/1.25 NA silicon objective (Olympus) and 488nm for GFP excitation and 561 for tRFP or mCherry or ii) on an upright Nikon E1000 microscope coupled to a Yokogawa CSU10 spinning disc confocal scanner unit with a 20x/0.75 NA air objective (Nikon) or 10x/0.5 NA air objective (Nikon) and illuminated using a Spectral Applied Research LMM5 laser module (491 nm for GFP excitation; 561 nm for Ruby or TagRFP or mCherry). Confocal stacks using a 2μm z-spacing were acquired every 20-40 s.

Laser wounding was performed on a two-photon scanning miscroscope (LaVision Biotec TriM Scope II). A tunable ultrafast laser (Insight DeepSee, SpectraPhysics) was tuned to 930 nm and the laser power adjusted to approximately 500mW. A square region of interest (ROI) of ∼40μm in width was defined in one focal plane followed by single laser scan across the ROI at a pixel spacing of 240nm and dwell time of 13 μs. Confocal stacks were acquired immediately after, using a 25x/1.05 NA water-dipping lens. GFP was imaged with 930nm and DsRed was imaged with a 1040nm line. For imaging 5-LO translocation, the resolution of imaging with the two-photon microscope was limiting. Larvae were thus transferred (within 10-20 min) for imaging onto an upright Nikon E1000 microscope coupled to a Yokogawa CSU10 spinning disc confocal scanner unit with a 40x/0.80W water objective (Nikon). In some cases, PI (50μg/ml) was added to the medium 30 min prior to imaging. PI penetration was observed only with superficial laser wound.

For the Cxcl8a response assay, HEK293T cells were cultured in DMEM (Invitrogen) containing 10% FBS (GIBCO ThermoFisher Scientific) and 1% Penicillin/Streptomycin (Sigma). HEK293 cells were transfected with Cxcl8a-mCherry using Lipofectamine-2000 (Invitrogen) (construct described in [30]. Transfected cells were incubated at 37°C (with 5% CO₂) overnight, harvested the following morning and resuspended in DPBS (Invitrogen) at a density of 30x10⁶/ml. Cells were transplanted above the yolk into 48hpf Tg(lyz:GCamp6F) larvae as previously described [30]. Validation of Cxcl8a-mCherry secretion and function in vivo was described previously [30].

Tail fin wound infection experiments

3dpf larvae were anesthetized with 160-200mg/L MS-222 (Sigma) and their tail fin was amputated using a sterile surgical scalpel blade (Swann-Morton, 23). Larvae were subsequently (within 5 min) incubated for 2 hours at 33°C in Ringer (145mM NaCl, 2mM KCl, 1.5mM K₂HPO₄, 1mM MgSO₄, 10mM HEPES, 2mM CaCl₂ and 10mM glucose, pH7.2) medium with 1.10⁷/mL P. aeruginosa PAO1 strain (provided by Dr. Martin Welch). Following incubation, larvae were washed 5 times in PBS and separated in individual wells containing 100μL of PBS on a 96 well plate, to avoid transmission of infection across larvae. Survival and bacterial burden were monitored at 2, 6 and 18 hours post-wounding. For determining the bacterial burden, larvae were homogenized with a pestle gun (Anachem LTD) in 100μL PBS in a 4.5mL Eppendorf tube. Serial dilutions of these homogenates were then plated on P. aeruginosa isolation agar (Scientific laboratory supplies LTD) supplemented with cetrimide (200 mg/L) and nalidixic acid (15 mg/L) (CN supplement; Scientific laboratory supplies LTD) and incubated for 24h at 37°C. Colonies were counted to determine the number of colony forming units contained in one fish.

Two-photon imaging of wound colonization by bacteria

P. aeruginosa PAO1 strain were incubated for 30 min with 10μM Syto62 (Invitrogen) in PBS at 37°C. This was followed by four washes by centrifugation in PBS. Larvae were anesthetized with 160-200mg/L MS-222 (Sigma) at 3dpf and mounted in 1% agarose in a custom-made imaging chamber, consisting of a round coverslip above and below the sample sealed onto a metallic ring. Agarose was allowed to set and then was removed around the tail with a glass capillary in order to allow bacterial swimming toward the wound. E3 medium with 1.10⁷/mL labeled PAO1 and 160-200mg/L MS-222 was added to the chamber, which in turn was sealed and sterilized on the outside using a 1% virkon solution and 70% ethanol. The fish were subsequently imaged and wounded on a two-photon laser scanning microscope as described above.

Chemoattractant injections in the otic cavity

Morpholino injected and non-morpholino injected 3-dpf larvae were injected in the otic vesicle with 1nL of 30nM LTB4 (Sigma),30nM Cxcl8a (ProteinTech, custom-made) or 200μM non-hydrolysable ATP-γS (Sigma) in combination with 10% phenol red to track the injection. The control solution for LTB4 was ethanol at 1:1000 dilution. The control solution of Cxcl8a and non-hydrolysable ATP-γS was 0.1% BSA in PBS. Larvae were imaged on a spinning-disc confocal microscope 15 min after injection or fixed for Sudan Black staining 45 min after injection.

Whole-mount Sudan Black staining and immunohistochemistry

Tail-fin amputated larvae or larvae microinjected in the otic cavity were fixed after 3 hours and 45 min respectively in 1ml of 4% ethanol-free formaldehyde (Polysciences, Warrington, PA) in PBS (PBS; Sigma-Aldrich) overnight at 4°C with agitation. Fixed larvae were rinsed in PBT (PBS with 0.1% Tween-20; Sigma-Aldrich) twice for 5 minutes and incubated in 1ml Sudan Black (Sigma-Aldrich) for 15 min. Following staining, larvae were washed in 70% ethanol for several hours and transferred to 30% ethanol overnight at 4°C with agitation. Larvae were washed in PBT for 10 minutes, transferred to increasing concentrations of glycerol and stored in 80% glycerol at 4°C. Larvae were imaged on an optical microscope Stemi 2000-CS (ZEISS) mounted with axiocam ERcs 5 s (Zeiss).

For immunohistochemistry, 3dpf larvae were fixed overnight in 4% formaldehyde (ThermoFisher) at 4°C, washed twice in PBT (PBS, 0.1% Tween-20) and then dehydrated overnight in methanol 100% at −20°C. Larvae were progressively rehydrated by decreasing methanol concentration, heated for 15min at 70°C and fixed in ice cold acetone for 20 min at −20°C. After blocking in 10% sheep serum (Sigma-Aldrich), proteins were probed with primary antibody and revealed using secondary antibody. GFP⁺ neutrophils, tRFP-5-LO and Cx43 were detected using chicken anti-GFP (abcam) at 1:500, rabbit anti-tRFP (Evrogen) at 1:500 and rabbit anti-Cx43 (Sigma) at 1:300 respectively. Secondary antibodies used were anti-chicken-Alexa488 (Invitrogen) and anti-rabbit-Cy3 (Jackson) at 1:500. Nuclei were stained with DAPI at 0.5μg/mL (Sigma Aldrich). Subsequently larvae were embedded in 80% glycerol and mounted for confocal observation (Olympus Fluoview FV1000).

RT-PCR of Cx43 genes in neutrophils

RNA extraction: Larvae were snap frozen in liquid nitrogen after removal of E3 medium. RNA was extracted with the RNAeasy minikit (QIAgen) according to manufacturer’s instruction. RNA was then reverse transcribed with SuperScript III Reverse Transcriptase (Invitrogen). PCR was performed using the KOD Hot Start DNA polymerase kit (Novagen, TOYOBO).

• List of primers used for RT PCR:

• cx43 Forward 5′-GCTCTCCA CTCTTTACTTCTTTCCAG-3′

• cx43 Reverse 5′-GTATTGCACTTGAAAGCTGACTGC-3′

• cx43.4 Forward 5′-GAGTCGTCATCGCGAGACATTGA-3′

• cx43.4 Reverse 5′-GTCTATGAGTCTCAATCAAGCATGGATCC-3′

• lta4h Forward 5′-TCTGAGAAGGAATATGTGGATGAA-3′

• lta4h Reverse 5′-CAGCAAGAGATCTGTCTCCA-3′

Extraction of cell trajectories

Analysis of neutrophil trajectories was performed in Imaris v8.2 (Bitplane AG, Zürich, Switzerland) on 2D maximum intensity projections of the 4D time-lapse videos. For laser-wounded larvae, unless otherwise indicated, trajectories were extracted from a cropped area covering the ventral fin and the part of CHT in which there was neutrophil immobilization. For ventral fin-wounded larvae trajectories were extracted from a cropped area covering the entire ventral fin. A track duration threshold of 3 time-frames was defined to exclude short-lived tracks. Manual track corrections were also applied where needed. Instantaneous (x,y,t) coordinates over time were exported into Microsoft Excel 2016 spreadsheets files (Microsoft Corporation, Redmond, WA).

Extraction of cell surface data

Analysis of neutrophil cluster size and calcium signal was performed in Imaris. Neutrophils were segmented as surfaces and manual surface splitting or merging was applied where needed. Instantaneous neutrophil cluster size and calcium signal intensity were exported into Microsoft Excel 2016 spreadsheets files.

Definition of wound perimeter and of clustering versus migrating cells

For laser wounds, the perimeter of the wound was manually defined in MATLAB R2018b (The MathWorks, Natick, MA) as a set of points surrounding the autofluorescent area of the wound. For mechanical ventral fin wounds, the perimeter of the wound was manually defined in MATLAB as a set of points surrounding the area maximally occupied by neutrophils at the wound (occupied wound area), as indicated by a continuous surface of high intensity on a time-projection of the movie, as previously described [65].

Quantification of GCamp6F levels

Except for Figure 2 (see previous section on 5-LO-translocation analysis), calcium signal values were extracted from Imaris and imported into MATLAB for plotting. For the laser wound experiments, calcium values for individual segmented neutrophils were normalized to the mean calcium value of the neutrophils in the whole area outside the wound, prior to wounding. In all corresponding bar graphs, the first 3-9 frames post-wounding were excluded to eliminate distortion of the neutrophil data by the tissue-scale calcium wave. For mechanical fin wound videos, the calcium values for individual neutrophils were normalized to the mean calcium value of the neutrophils in the whole area outside the wound, at the first time point of imaging. For the analysis of GCamp6F levels in the ear, the mean intensity of cells in the ear was divided by the mean intensity in cells outside the ear.

Quantification of GCamp6F levels with neutrophil cluster size

Neutrophil cluster size and calcium signal values were computed in MATLAB and plotted against each other. A threshold of 60 pixels on the size of detected objects was applied to eliminate false detections.

Calculation of neutrophil radial speed

Radial speed was calculated in MATLAB using the following Equation [8]:

$$u_r=u\times cos\theta$$

where u is the instantaneous speed of neutrophil between two successive positions and θ is the angle between the vector of the movement and the vector that connects the position with the wound. The angle θ was calculated using the vector between the neutrophil position (centroid) and its nearest point to the wound. When the cosine of θ has value 1, the neutrophil migrates directly toward the wound while when it has value −1, the neutrophil migrates directly away from the wound. To uncover trends in directionality of motion in embryos independently from intra-embryonic variation in speed levels, we used a normalization. Instantaneous speed values for individual neutrophils were divided by the mean instantaneous speed value of the corresponding embryo. Normalized radial speeds were computed with the equation:

$$u{r}_{norm}=u_{norm}\times cos\theta$$

Normalized radial speed values were binned every 5 or 7.5 min. For the laser wound experiments, the first 3-9 frames post-wounding were excluded for consistency with the calcium signal calculations.

Detection and scoring of 5-LO translocation in zebrafish neutrophils

Automated detection [56] was not applicable to neutrophils due to the irregular shape of the nucleus and their dynamic movement. We thus used visual inspection of the time-lapse videos on Fiji [66] and representative sample videos were confirmed by two viewers. Only unambiguous translocation events were scored.

Analysis of 5-LO translocation in relation to distance or GCamp6F intensity

Frames in which 5-LO translocation events were detected and thereafter analyzed with MATLAB in an automated fashion. Individual cells were segmented using marker-based watershed segmentation and intensity thresholding. Mean fluorescence intensities of the GCamp6F signal in segmented neutrophils were subsequently computed. For each neutrophil, the fluorescence intensity was normalized to the most fluorescent cell in the corresponding frame to allow pooling of values across embryos with different imaging settings. The wound center was manually inputted and the distance of individual neutrophil centroids from the wound center was automatically computed using a custom MATLAB script.

Analysis of GCamp6F in neutrophil cell-cell contacts and neutrophil contacts with necrotic tissue

Contacts between bright and dim GCamp6F⁺ neutrophils or contacts with PI⁺ necrotic cells were counted and classified according to whether a sharp increase of fluorescence was observed in the dim cell upon contact or not. These events were quantified using visual inspection of the time-lapse videos on Fiji. Only unambiguous events were scored.

Evolution of neutrophil GCamp6F levels and speed over time and upon contact with necrotic cells

Individual neutrophils were visually inspected to determine the time-point that they touched the PI-stained necrotic cells. This time-point was considered as the time-point 0. The neutrophils were tracked for 180 s before and after this time-point. Individual neutrophil calcium values were normalized with the calcium value of the first time-point of the track.

Analysis of fluorescence intensity of bacteria over time

Segmentation of bacteria was done in Imaris using surface segmentation in the 3D dataset. The segmentation was manually limited to a volume spanning the z dimension of the wound core, taking also into consideration the wound autofluorescence and the location of the neutrophil cluster. Corrections in tissue drift were performed by changing the tracking area in all 3 dimensions (x,y,z) in time, to allow accurate segmentation of this area. Intensity values were extracted in Imaris. A custom-written MATLAB script was used to normalize values to the fluorescent intensity of all bacteria in the first frame before wounding, and to the maximum initial fluorescent intensity of bacteria in the wound area (usually between 5-10 min post wounding), and plot the mean of different time bins across independent larvae. A contribution of fluorescence photobleaching was excluded by performing similar computations in a region of interest with bacteria outside the larva.

Quantification and Statistical Analysis

All error bars indicate SEM. All p values were calculated with two-tailed statistical tests and 95% confidence intervals. t test (pairwise comparisons) and one-way ANOVA (multiple group comparisons) were performed after distribution was tested for normality otherwise non-parametric tests were performed (Mann-Whitney for two-way comparisons and Kruskal-Wallis with Dunn’s post-test for multiple comparisons). Unless otherwise indicated, tests were unpaired. Statistical tests were performed in Prism8 (GraphPad Software, La Jolla, CA). The statistical test and the n number are indicated in the figure legends. The error bars show standard error of the mean across individual embryos or cells except for Figures 2E and 2F where error bars represent 95% confidence intervals of the median. Where the distribution was verified as normal, outliers were removed by applying Rout test. Live imaging experiments were acquired in minimum three independent experiments. In figure panels, ^∗ corresponds to p < 0.03, ^∗∗ to p < 0.002 and ^∗∗∗ to p < 0.0002. In line graphs with binned data, ^∗ is indicated for the bin with smallest detectable significant difference.

Published: June 4, 2020