Drosophila immune cells extravasate from wing veins to sites of tissue damage

Using the distal-most region of 75-h APF pupal wings, we generated small 40-µm-diameter laser wounds within the wing epithelium at ~40 µm from vein L3 (Fig. 2, A and B). Strikingly, this injury reproducibly triggered local hemocyte extravasation from the vessels into the damaged tissue (Fig. 2 B and Video 6). Within 30 min of injury, hemocytes locally halted their flow and began interacting with the vessel wall, extending pseudopodial-like probes along and through the luminal surface as though scanning for optimal exit sites (Fig. 2, B and E; and Video 6; mean 48.7 ± 17% of hemocytes passing the wound site, n > 300, and 77% of videos examined, n > 30). Similar probing behavior is highly characteristic of vertebrate leukocytes that extend protrusions through junctions between adjacent endothelial cells or into the endothelial cell body in order to detect chemotactic gradients associated with the endothelium or subendothelial space (Carman et al., 2007; Nourshargh and Alon, 2014). In the majority of wounds examined (65% of videos, n > 30, up to 6 h after injury), hemocyte probing led to at least one hemocyte successfully extravasating through the vessel wall and migrating across the extravascular space to the site of damage (Fig. 2 B). Extravasating hemocytes underwent a series of dramatic morphological changes during vessel exit, generating a dynamic protrusive leading edge in the direction of migration and a rear contractile uropod during tail retraction (arrows, Fig. 2 B).

Using specialist IMARIS software, which offers superior spatial resolution than traditional z-stack maximum intensity projections, we visualized the 4D time-lapse imaging data as dynamic 3D videos (Fig. 2 C and Video 7). By 2 h after wounding, multiple hemocytes had extravasated from the vessel (on average 10.3% of hemocytes passing the wound site, n > 300), and on average 70.2% (n > 35) of these extravasated hemocytes reached the injury site (Fig. 2 Civ; mean of 2.1 hemocytes, SD = 1.1, n = 29 videos), where they often remained for several hours phagocytosing necrotic wound debris (Fig. 2 Cvii) until the wound had reepithelialized. However, occasional emigrated hemocytes (34.2% on average, n > 25) reverse transmigrated back toward, and into, the vessel (Fig. 2, Civ–Cvi; see color-coded asterisks of reverse migrating hemocytes), often at the same site from which they first emerged, as previously reported for neutrophil reverse migration in zebrafish and mice (Mathias et al., 2006; Tauzin et al., 2014; Colom et al., 2015; Ellett et al., 2015; Nourshargh et al., 2016; Wang et al., 2017). While the majority of extravasation is complete by 2 h after wounding, long-term imaging revealed that hemocytes could be recruited for up to 10 h after the completion of wound closure (data not shown). Intriguingly, hemocyte trajectory analysis suggests the existence of particular “hot spots” of extravasation, with the majority of hemocytes leaving the vessel immediately adjacent to the wound and several hemocytes streaming out, one after another (Fig. 2, D and D’), although hemocytes could also extravasate from the vessel up to 100 µm up- or downstream of the wound. Following extravasation, hemocytes migrated with a mean velocity of 3.8 µm/min (SD = 0.7) through the extravascular space to the injury site, similar to our observations for hemocytes migrating to pupal wounds at stages before vessel formation (Weavers et al., 2016b). We observed only rarely hemocytes in this wing vessel region (Fig. 1 I) extravasating from the vein in unwounded conditions (in 6.7% of videos examined, a single hemocyte extravasated, n = 30), perhaps triggered by local apoptosis in the intervein epithelium. Although these extravascular hemocytes are responsive to epithelial wounds (data not shown), they were excluded from our quantification if they had extravasated before wounding.

Correlative light-electron microscopy (CLEM) captures specific ultrastructural details during hemocyte extravasation

Live imaging of wounded pupal wings allowed us to pinpoint precise moments when hemocytes extravasate from vessels (Fig. 2). To capture ultrastructural details of these events, we fixed individual pupae at key moments to enable CLEM, a technique previously used to analyze in vitro intracellular trafficking (de Boer et al., 2015). In this way, we captured single time points that enable us to study (i) hemocytes in the act of extravasation (h-ex; Fig. 3, A–C and F–J); (ii) extravasated hemocytes en route to the wound (Fig. 3 L); as well as (iii) hemocytes at the wound actively clearing cell debris (h-w; Fig. 3, A–C, N, and O). Low-magnification transverse sections demonstrate the route of hemocyte extravasation from the vein lumen, through the interstitium to the site of tissue damage (resin histology, Fig. 3 D, and TEM, Fig. 3 E), while high-magnification serial TEM sections reveal cellular through to ultrastructural details of hemocyte extravasation (Fig. 3, F–K; and Fig. S2, A and B) and post-extravasation migration toward the wound (Fig. 3, L–O; and Fig. S2, E–H).

Open in a separate window

Figure 3.

CLEM of hemocyte extravasation. (A) In vivo imaging of extravasation (hemocytes, red, srp>mch-moesin; and epithelium, green, ubiquitous GFP-Moesin). (A–C) Samples fixed at key moments (hemocyte, h-ex, initiating transmigration; and hemocyte, h-w, reached the wound). Arrows indicate hemocytes in the process of extravasation (h-ex), and arrowheads indicate extravasated hemocytes at wound (h-w). (D–O) Samples prepared for TEM. (D and E) Resin transverse section (D, methylene blue) and TEM section (E) illustrate the route from vein to wound with a hemocyte (h-ex, false-colored red) traversing the vein wall (E). (F and G) TEM reveals hemocyte probing of the vessel wall, with long (F) and short (G) protrusions targeting junctions between adjacent vein wall cells (arrow, G). Arrows (black, F) indicate the end of the long hemocyte probing protrusion. White arrow (F, inset) indicates intact vein wall junction. (H–K) Prior to extravasation, the vein wall barrier is intact (H; arrows, vein wall junctions), but extravasating hemocytes move across and flatten this vein wall barrier (arrowheads, I and J). Extravasating hemocytes are associated with large darkly staining deposits (arrows, I), and vein cells exhibit vacuolation (arrows, K). (L–O) Hemocytes use long protrusions to migrate through the extravascular space toward the wound (arrow, L). Wounds are repaired by long epithelial tongues (arrows, M and M'). Hemocytes at the wound site (magenta, N and O) clear up necrotic wound debris (yellow, indicated by arrows). All images were taken from two 75-h APF Drosophila pupae prepared for CLEM. Also see Fig. S2.

Extravasating hemocytes (h-p) extend probing protrusions toward the vein wall (Fig. 3 F) that target the intercellular junctions between adjacent vein wall cells (arrow, Fig. 3 G). Drosophila vein wall cells are linked by tight intercellular adherens junctions (Fig. 3 H and Fig. S2, A and A’; O’Keefe et al., 2007), as in vertebrate vessels (Muller, 2003), but, at sites of hemocyte extravasation, we observed dramatic ultrastructural changes (Fig. 3, I and J; and Fig. S2 B). Vein wall cell contacts are interrupted (Fig. 3 I and Fig. S2 B), and we observed long vein cell “tongues” (arrowheads, Fig. 3, I and J), suggesting that the vein cell barrier might have been flattened by extravasating hemocytes; this is consistent with live-imaging observations of dextran leakage from the wing veins following wounding (Fig. S2 C). Loosening of junctions is also characteristic of vertebrate immune cell extravasation (Muller, 2011), as vertebrate immune cells generate gaps of micron-scale size when squeezing through inflamed endothelial barriers in vitro and in vivo (Shulman et al., 2009). Extravasating hemocytes are closely associated with numerous darkly staining dense deposits (arrows, Fig. 3 I; and Fig. S2 D) that were also observed inside (and in the vicinity of) hemocytes in unwounded conditions (Fig. 1, J and K), which we speculate might reflect mechanisms that facilitate exit from the vein. The vein cells themselves became highly vacuolated (Fig. 3 K), reminiscent of the vacuoles observed in endothelial cells during vertebrate immune cell diapedesis (Muller, 2015, 2016).

Following extravasation, hemocytes extend long leading-edge protrusions through small gaps (arrows, Fig. 3, L and L’) between upper and lower intervein epithelia as they migrate toward the wound and exhibit slender retracting uropods at their rear (Fig. S2 F). This matrix-rich, extracellular space between intervein epithelia is present even in unwounded tissue (Fig. S2 E; Fristrom et al., 1993). Less organized patches of extracellular material were observed in advance of extravasated hemocytes (Fig. S2 G), suggesting that hemocytes might remodel material in the interstitial space as they migrate toward the wound. Similar material is sometimes observed within extravasated hemocytes (Fig. S2 H). Interestingly, we found that levels of the secreted Matrix Metalloproteinase 1 (MMP1; Stevens and Page-McCaw, 2012) increased following wounding (Fig. S2, I–K), and we observed the appearance of punctae of Collagen IV (Drosophila Viking) within hemocytes (Fig. S2 L), although it is unclear whether this reflects hemocyte synthesis or internalization of Collagen IV. Epithelial wound repair occurs in parallel with the inflammatory response, and we observed wound edge epithelial tongues extending toward one another to seal the wound gap (Fig. 3 M). On reaching the wound, hemocytes play an active role in the repair process by phagocytic clearance of necrotic debris (arrows, Fig. 3, N and O).

Drosophila hemocyte extravasation from veins is integrin dependent

The arrest of leukocytes on the luminal endothelium surface at the site of vertebrate inflammation requires activation of at least one of the major leukocyte integrins (Dustin and Springer, 1988; Elices et al., 1990; Hynes, 1992; Schenkel et al., 2004; Phillipson et al., 2006; Herter and Zarbock, 2013). We therefore tested whether integrins play a conserved role during Drosophila immune cell extravasation. Drosophila possess five integrin α-subunits (αPS1–5) and two integrin β-subunits (βPS and βν; Bulgakova et al., 2012); the αPS2 integrin inflated is known to be required for the invasive developmental migration of embryonic hemocytes into the tail (Siekhaus et al., 2010) and early pupal hemocyte motility (Moreira et al., 2013). We found that RNAi-mediated knockdown of inflated integrin (using multiple independent RNAi lines) within late pupal hemocytes (using the hemocyte-specific srp-Gal4 driver) caused dramatic defects in hemocyte extravasation (Fig. 4, A–C; Fig. S3, A and B; and Video 8). inflated-RNAi hemocytes rarely arrest and extend probing protrusions across the vessel wall adjacent to the injured tissue (Fig. 4 C and Fig. S3 A; imaging for 3 h after injury). Consequently, significantly fewer inflated-RNAi hemocytes successfully extravasate from the vessels (Fig. 4 C; extravasation observed in 6.7% or 13.3% of videos for inflated-RNAi TRiP.JF02695 or TRiP.HMC06096, respectively, versus 38% of controls over short-term 3-h imaging). We saw no significant defect in the movement of hemocytes within pupal wing veins before (or after) wounding following inflated-RNAi (Fig. S3 B), suggesting that basal hemocyte motility within veins of 75-h AFP pupal wings does not require inflated. Hemocytes lacking inflated, however, migrated more slowly toward the wound following extravasation, suggesting that post-extravasation migration does require inflated (Fig. S3 B).

Open in a separate window

Figure 4.

Hemocyte extravasation requires integrins and the GPCR Tre1. (A–C) IMARIS visualization of control (A) and hemocyte-specific inflated-RNAi^{P{TRiP.JF02695}} hemocytes (B, using srp-Gal4) in 75-h APF pupal wings. Unlike controls (red, arrowheads, A and A’), hemocytes with inflated-RNAi^{P{TRiP.JF02695}} exhibit defects in vessel extravasation (arrowheads, B and B’, and quantified in C). (D–H) Hemocytes with tre1-RNAi^{P{TRiP.HMS00599}} (using srp-Gal4) become significantly elongated and branched during extravasation (arrowheads, Dii and Diii, and morphology quantification, E–H) and fail to retract their vessel-bound tails (arrows, Div). Arrowhead in Di indicates hemocyte probing of the vessel wall. Arrows in Diii indicate hemocyte tails that are retained within the lumen of the vessel. (I) Hemocytes with activated Rho1 (srp>Rho^V14) exhibit similar extravasation defects, becoming significantly elongated (G and arrowheads, I) and failing to retract their tails from the vessel (arrows, I). Asterisks (white) indicate individual hemocytes. (C, G, and H) Data are represented as mean ± SEM; ns, not significant; *P < 0.05, **P < 0.01, and ***P < 0.001 via multiple t tests (C) or one-way ANOVA (G and H), both with Sidak’s correction for multiple comparisons. pw, post-wounding. Also see Fig. S3 and Videos 8 and 9.

Time-lapse microscopy and wounding

White prepupae (0 h APF) were collected and maintained at either 25°C or 29°C until the required stage. For RNAi-mediated knockdown experiments, crosses were maintained at 18°C to minimize Gal4-mediated expression, and newly selected 0-h APF pupae then shifted to 25°C or 29°C to boost Gal4 activity. For imaging, pupae were dissected from their cases and mounted on glass-bottomed dishes (MatTek) using heptane glue. 75-h APF pupal wings were wounded using a nitrogen-pumped ablation laser (Spectra-Physics) attached to a Zeiss Axioplan 2 widefield imaging system. For microinjections, Cascade Blue Dextran (3,000 mol wt, final concentration of 5 mg/ml), Fluoresbrite Polychromatic Red Microspheres (1 µm diameter, 1:10 dilution), or Fluoresbrite YG Microspheres (0.5 µm diameter, 1:10 dilution) were injected into the thorax using a FemtoJet micromanipulator (Eppendorf). Imaging was performed on a Leica TCS SP5-II confocal laser-scanning microscope or light-sheet microscope (Lightsheet Z.1, Zeiss). Image preparation and analysis were performed using Volocity (PerkinElmer) and ImageJ (National Institutes of Health) software; bead velocities were quantified using the automated tracking feature within Volocity. Statistical analysis was performed using Prism software. For 3D reconstructions, imaging data were processed using IMARIS software (Bitplane). Figures were prepared using Adobe Photoshop, Adobe Illustrator, or Adobe InDesign software. Vein lumen position was determined by fluorescent dextran injection or careful examination of GFP-tagged Moesin labeling in individual z-sections.

Analysis of hemocyte extravasation and morphology

Hemocyte extravasation duration was defined as the time taken from the initial breaching of the vessel wall to when the cell body had completely moved across the vein wall into the intervein region. To quantify hemocyte morphology during extravasation events, maximum intensity Z-projections were generated from 4D imaging data to produce a single-plane time series. All images were convolved with a Gaussian kernel (σ = 1 pixel) to remove high-frequency noise. The series was binarized using a single intensity threshold calculated for the entire stack using Huang’s algorithm (Huang and Wang, 1995). Individual contiguous regions were manually selected for analysis and skeletonized using ImageJ (Schindelin et al., 2012; Schneider et al., 2012). Branches containing fewer pixels than a user-defined threshold were removed from further analysis. The remaining skeleton was analyzed using the Analyze Skeleton ImageJ plugin (Arganda-Carreras et al., 2010).

Immunostaining and in situ hybridization

Pupae were dissected from pupal cases and wounded, as above, before fixation for 10 min in a 1:1 mix of 8% paraformaldehyde with heptane. Pupal wings were removed and further fixed for 30 min in 4% paraformaldehyde before washing in 1× PBS/0.1% Triton-X. For immunostaining, pupal wings were blocked in 1% BSA in PBS and then incubated in primary antibody overnight at 4°C. Primary antibodies were detected using appropriate AlexaFluor488-, AlexaFluor568-, or AlexFluor647-conjugated secondary antibodies (Molecular Probes). Pupal wings were mounted on a glass slide in Vectashield (Vector Labs), and imaging was performed on a Leica SP5 confocal microscope. Immunostaining was performed with the following antibodies: anti-MMP1 (mix of 3A6B4, 3B8B12, and 5H7B11, all 1:10; Developmental Studies Hybridoma Bank), anti-GFP (1:500; Abcam), and anti-RFP (1:500; MBL). In situ hybridization was performed using standard techniques (Wilk et al., 2010), and tre1 transcript was detected using an antisense RNA probe transcribed from the tre1 cDNA clone RE07751 (BDGP Drosophila Genomics Resource Center).

RNA isolation, reverse transcription, and real-time quantitative PCR

RNA was isolated from control and srp-Gal4>UAS-tre1-RNAi(TRiP.HMS00599) 75-h APF pupae by crushing in TRIzol (Life Technologies) and RNA purified using an RNeasy Mini Kit (Qiagen). Equal quantities of RNA were then reverse transcribed using a Thermo Scientific Maxima First Strand cDNA Synthesis Kit, and genomic DNA was eliminated using double strand-specific DNase (Thermo Scientific). Relative quantification of gene expression was performed using SYBR Green Supermix with a real-time PCR machine. Tre1 gene expression was normalized to the expression of the housekeeping reference gene α-tubulin84B using the ΔΔCt analysis method. The following primers were used in this study: Tre1 F-primer 5′-ATTAGTGCCTGTGTCTTTGTGAC-3′ and R-primer 5′-GGAGATGCTTAGCGAAATGACG-3′; and α-tubulin84B F-primer 5′-CACACCACCCTGGAGCATTC-3′ and R-primer 5′-CCAATCAGACGGTTCAGGTTG-3′.

CLEM

75-h APF pupae were prepared for confocal microscopy as above. At appropriate time points, pupae were transferred to a glass vial for fixation (2% paraformaldehyde and 1.5% glutaraldehyde in 0.1 M sodium cacodylate buffer 1:1 with an equal volume of heptane) for 1 h at room temperature. Samples were washed three times (0.1 M sodium cacodylate) before transfer to distilled H₂O for removal of the cuticle and wing dissection. Wings were fixed further for 2 h (1% osmium in 0.1 M sodium cacodylate) before washing in 0.1 M sodium cacodylate and in distilled H₂O. Samples were dehydrated (30 min in 70%, 90%, 96%, and 100% ethanol) and incubated in propylene oxide (PPO; 3 × 20 min). PPO was replaced with a 50:50 mix of PPO:epon, incubated overnight, and then evaporated off for 2 h. Wings were transferred to fresh epon (3 g TAAB 812 Resin, 2 g dodecenyl succinic anhydride, 1.25 g methyl nadic anhydride, and 0.1875 g benzyl dimethylamine) for 24 h and then embedded/polymerized at 60°C for 72 h. Sections were cut on an Ultramicrotome (Leica EM UC6) and imaged using a Tecnai 12-FEI 120-kV BioTwin Spirit Transmission Electron Microscope with a FEI Eagle 4k × 4k charge-coupled device camera.

We would like to thank members of the Martin, Nobes, Richardson, and Wood laboratories for helpful discussion (in particular Andrew Davidson for advice on Rho signaling); Stephen Cross for developing new image analysis software for quantifying hemocyte morphology; and Jim Middleton (University of Bristol, UK) for expert advice on vertebrate extravasation. We also thank the Wolfson Bioimaging Facility (University of Bristol, Bristol, UK), Bloomington Drosophila Stock Center (University of Indiana, Bloomington, IN), and Vienna Drosophila Resource Center for Drosophila stocks.

This work was supported by a Wellcome Trust PhD studentship to L. Thuma, a Medical Research Council Project Grant to P. Martin and W. Wood (MR/J002577/1), and a Wellcome Trust Investigator Award to P. Martin (WT097791/Z/11/Z).

The authors declare no competing financial interests.

Author contributions: L. Thuma and H. Weavers performed the in vivo experiments. D. Carter assisted with the implementation of CLEM, performing sample sectioning and operating the transmission electron microscope. H. Weavers and P. Martin designed the study, coordinated the project, and wrote the manuscript. L. Thuma contributed to the design of the study and proofread the manuscript.